# Agent settings
Source: https://docs.ruleinside.com/agent-settings
Every setting in the agent editor: what it does, the values it accepts, its default and the edit level it needs.
The agent editor is where you shape one agent: what it says first, how it sounds and listens, how long calls last,
its instructions and, if it is switched on for your account, where it hands callers over to a person. To open it,
click **Agents** in the sidebar (see [Managing your agents](/managing-agents)).
We suggest you start with a handful of settings: the greeting in [First message](#first-message), the
[Voice](#voice), the [Language](#language), [Turn taking](#turn-taking) when the agent feels slow or talks over
callers, and [Call limits](#call-limits). You can usually leave the rest as it is. Under many of the tables below,
a **Recommended** note says when it is worth changing a setting.
The editor has two columns:
* **Left:** **First message**, **Personalization variables**, **Transfer to a person**, **Conversation** and
**System prompt**.
* **Right:** **Configuration** (**Name**, **Voice**, **Language**, **Language model**, **Turn taking**, **Call
limits**, **Conversation behaviour**), then **Knowledge base** and **Tools**. Each Configuration row except Name
opens a side panel.
## Who can change a setting
You can change a setting only when **both** of these are true:
1. You are an **owner** of the account. Members can open every agent and read every setting, but all fields are
greyed out for them and they see: "Only account owners can change agents, ask an owner of your team to make
edits." See [Who can do what](/roles).
2. The agent's **edit level** allows that setting. A setting above the agent's level shows a padlock or is greyed
out.
### Edit levels
Every agent has one of four edit levels. The level is shown as a badge in the editor header and under the agent's
name in the agents list. Each level includes everything in the levels above it in this table.
| Level | What you can change |
| - | - |
| **View only** | Nothing. A banner says: "Your plan is view-only. Contact us to change your agent, or upgrade for self-service editing." The save bar never appears. |
| **Name, voice & greeting** | **Name**, the choice of **Voice**, the **First message** text and **Personalization variables**. |
| **Instructions** | **Language**; **Creativity (temperature)**; **Stability**, **Speed**, **Similarity** and **Latency optimisation**; everything in **Turn taking**, **Call limits** (except voicemail) and **Conversation behaviour**; **Store timezone**; **Transfer to a person**; **Additional instructions**; **Who speaks first** and the wait for the caller; **Recording notice**. |
| **Full control** | **TTS model family**, **Audio output format**, **Text normalisation**, **Max response length (tokens)**, the whole **System prompt** (it takes the place of Additional instructions), the **System tools** switches, and all changes on the [Tools](/tools) and [Knowledge](/knowledge) pages, including **Configure RAG**. |
* Agents you create yourself (blank or cloned) always start at **Full control**.
* Your main agent, the one RuleInside set up for you, keeps the level RuleInside set for it. It carries the
**Managed** badge, which only marks it as your main agent: the edit level, not the badge, decides what you can
change.
* Some settings can never be changed from the portal, at any level: the language model name and its reasoning
setting, the voicemail message, **Transfer to agent**, RuleInside's own store tools and knowledge documents, and
RuleInside's sections of the prompt. Contact RuleInside if you need one of these changed.
The settings that need **Full control** are the ones that can stop an agent working properly when set wrongly.
For example, a wrong audio output format breaks the sound on phone calls.
## Saving your changes
When you change something, a bar appears at the bottom of the page with two buttons:
* **Discard** puts every field back to the values loaded with the page.
* **Save changes** sends your changes (the button shows "Saving…" while it works). Every field is locked until
the save finishes.
What to know about saving:
* Only fields you changed, and are allowed to change, are sent. If there is nothing to send you see "No changes
to save."
* **Closing a side panel does not save it.** Changes in the Voice, Language, Language model, Turn taking, Call
limits and Conversation behaviour panels, the First message card, transfer numbers, time zone and prompt all
save together when you click **Save changes**.
* If a field is empty or out of range you see "Some settings need fixing before this can be saved. Open the
highlighted section to see what." Nothing is saved. Open the highlighted panel to find the message under the
field.
* On success you see "Agent updated." and the form reloads with the saved values. The change applies to the live
agent.
* If the save is refused you see the reason, for example "Not permitted at this agent's edit level (…): …", or
"Could not update the agent."
Two things do **not** use the save bar and save as soon as you click: **Personalization variables** (they have
their own **Save variables** button) and the **System tools** switches on the [Tools](/tools) page.
## First message
The **First message** card sets how the call opens. Its description reads "The first thing customers hear." or,
when the caller speaks first, "What the agent says after the caller answers."
| Field | What it does | Values and limits | Default | Edit level |
| - | - | - | - | - |
| **Who speaks first** | Who talks first when the call connects | **Agent speaks first**: "The agent says the greeting as soon as the call connects." **Caller speaks first**: the agent waits for the person to answer ("Hello?"), then says the greeting. | As set on the agent | Instructions |
| **Wait for the caller to speak (seconds)** | Shown only with **Caller speaks first**. How long the agent waits for the caller to speak first. | 1 to 30, or empty. Empty uses the reply wait from **Turn taking**. | Not set | Instructions |
| **Recording notice** | Where the agent says the call may be recorded | **In the opening line**: "The greeting itself says the call may be recorded." **After the first exchanges**: the agent says it once, in its second or third reply. If the person objects to being recorded, it says someone from the team will get back to them and ends the call. | As set on the agent | Instructions |
| **What the agent says** | Shown only with **After the first exchanges**. The recording sentence, said word for word. | Up to 500 characters. Required while this option is chosen. | Empty (placeholder "This call may be recorded for quality purposes.") | Instructions |
| First message text box | The greeting | Up to 1,000 characters. Required with **Agent speaks first**. | Placeholder "Hi! How can I help you today?" | Name, voice & greeting |
Recommended:
* **First message text:** keep it short, and say that the caller is speaking with an AI assistant and that the
call may be recorded (the portal's own hint under the box asks for both).
* **Who speaks first:** keep **Agent speaks first** for most stores, so callers hear the greeting straight away.
Choose **Caller speaks first** if you would rather the agent waited for the person to say "Hello?" first.
* **Wait for the caller to speak (seconds):** leave it empty, so it uses the reply wait from **Turn taking**. Set a
value only if you want a different wait at the very start of the call.
* **Recording notice:** keep **In the opening line** unless your greeting feels too long. Then choose **After the
first exchanges** and move the recording sentence into **What the agent says**.
* **What the agent says:** one short sentence is enough, for example "This call may be recorded for quality
purposes."
Good to know:
* The greeting should tell callers they are speaking with an AI assistant and that the call may be recorded. The
portal does not stop you saving a greeting without them, but always include them.
* With **After the first exchanges**, the greeting must still say the caller is talking to an AI assistant. Take
the recording sentence out of the greeting, or it will be said twice.
* **Wait for the caller to speak (seconds)** is the same setting as **Wait at the start of the call (seconds)**
under [Call limits](#call-limits). Changing one changes the other.
* **Caller speaks first** needs the **Instructions** level. At **Name, voice & greeting**, a greeting you type
makes the agent speak first.
* If **Who speaks first** and **Recording notice** are not shown on your agent, contact RuleInside.
## Personalization variables
Values your storefront passes in when a call starts, such as the customer's first name. Use one in the first
message or the instructions by writing its name in double curly braces, for example `{{first_name}}`. The default
is used when no value is passed in.
| Field or button | What it does | Values and limits |
| - | - | - |
| Name box | The variable's name (placeholder "name (e.g. first\_name)") | Starts with a letter; letters, numbers and underscores; up to 64 characters |
| Default box | The value used when none is passed in (placeholder "default (e.g. there)") | Up to 500 characters; longer text is cut |
| Trash button (**Remove variable**) | Removes the row from the form | |
| **Add variable** | Adds an empty row | Up to 30 variables |
| **Save variables** | Saves the whole list straight away ("Saving…" while it works) | |
* Edit level: **Name, voice & greeting**. The card is hidden on a **View only** agent. When you cannot edit, the
boxes are greyed out and the card says "This agent is managed, variables are configured by RuleInside."
* **Save variables** replaces your whole list. A row you removed is deleted when you save, and rows with an empty
name are dropped. On success: "Variables saved." On failure: the reason, or "Could not save variables."
* Some names are reserved by RuleInside: names starting with `system__` or `secret__`, and names that start and
end with an underscore. They are refused with "'…' is a reserved variable name." RuleInside's own variables are
never shown here and never removed by your save.
* When the list is empty the card shows "No variables yet."
## Transfer to a person
"The agent hands the caller over to one of your numbers when the situation calls for it. Works on phone calls
only."
This card appears when RuleInside has switched transfers on for your account. If transfers are not switched on
but RuleInside already set one up for this agent, the card shows: "RuleInside set this transfer up for you. You can
remove it here at any time. To add a number or change when the agent transfers, contact RuleInside support." In
that case the fields are read-only, there is no add button, and the **Remove** buttons still work.
Each transfer rule has:
| Field | What it does | Values and limits |
| - | - | - |
| **Phone number** | The number the caller is handed over to | International format: `+` then 7 to 15 digits, for example `+15551234567` |
| **When to transfer** | When the agent should hand over (placeholder "The caller asks to speak with a person, or the agent cannot help.") | Required, up to 500 characters |
| **Remove** (trash button) | Deletes this rule from the form | |
* **Add a number** adds a rule. You can have up to 5; the button is greyed out at 5.
* Edit level: **Instructions**, owners only. Rules save with **Save changes**.
* With no rules the card says "No transfer numbers. The agent never hands a call over." Removing every rule
switches transfers off.
* RuleInside keeps a record of every change to transfer numbers and who made it.
A number is refused when:
| Situation | Message you see |
| - | - |
| Your account has no active phone number yet | "Transfers need a phone number on this store first." |
| One of your store's numbers cannot be read | "This store has a phone number we cannot read (…); fix it under Phone numbers before setting a transfer." |
| It is your store's own number | "… is this store's own number; the agent would call itself." |
| It is not a valid number | "… is not a valid phone number." |
| It is not a mobile, landline or toll-free number (for example premium rate, shared cost or VoIP) | "… cannot be used for transfers. Use a mobile or landline number." |
| It is in a different country from your store's phone numbers | "… is in a different country from this store's phone number." |
| The same number is entered twice | "Each transfer number may appear only once." |
| Transfers are not switched on and you add or change a rule | "Transfer to a person is not enabled for this account. You can remove a transfer, but not add or change one." |
If the agent has a kind of transfer the portal cannot display, the card says "This agent has transfers that cannot
be shown or changed here. Contact RuleInside support to change them." and nothing can be changed.
## Conversation
The **Conversation** card is marked "Managed by RuleInside." and has a padlock, but one field in it is yours:
* **Store timezone** (edit level **Instructions**): a list of time zones such as Europe/Rome, plus UTC and **Not
set**. "Lets the agent tell callers the right local time, for example whether the shop is open now." Saves with
**Save changes**. Anything other than a real zone name is refused with "Timezone must be a real zone name, e.g.
Europe/Rome."
* **Variables** ("Identity & call context"): the variables RuleInside passes into every call. Shown for
information only; never editable.
## System prompt
What the **System prompt** card shows depends on the agent's edit level.
| Edit level | What you see | What you can do |
| - | - | - |
| **Full control** | "The full agent prompt. Platform sections must stay intact, saving without them is rejected." | Edit the whole prompt in one box, up to 20,000 characters. |
| **Instructions** | "Add your own guidance on top of the managed instructions. Platform behaviour (memory, store lookups, call handling) stays managed." | Write in **Additional instructions** (up to 8,000 characters; "Added to your agent's instructions. Leave blank to remove."). Click **Show** under **Current agent instructions (managed)** to read the full prompt. |
| **Name, voice & greeting** or **View only** | "How your agent behaves today." | Read the full prompt under **Current agent instructions (managed)**. |
* **Additional instructions** are added to the prompt in their own section; the rest of the prompt is never
touched. Saving the box empty removes that section.
* At **Full control**, if your agent's prompt contains RuleInside's platform sections, they must still be there
when you save. Otherwise the save is refused with "Your system prompt must keep its platform sections (…).
Restore them and save again." A blank agent you created has no platform sections.
* The opening greeting (when the caller speaks first) and the recording sentence (when said after the first
exchanges) are kept in their own sections. They are not shown in this box and are kept automatically when you
save the prompt.
## Name
**Name** (edit level **Name, voice & greeting**): 1 to 100 characters. An empty name is refused. The new name
also appears in the agent selector and the agents list.
## Voice
The **Voice** row shows the voice's picture, name and accent (or "Choose a voice"). Click it to open **Agent
voice**: "The voice your agent speaks with." The defaults below are what the panel shows when the agent has no
value set.
| Field | What it does | Values | Default | Edit level |
| - | - | - | - | - |
| Voice button (**Change voice**) | Opens the voice picker | Any voice in your voices or the library | | Name, voice & greeting |
| **TTS model family** | The speech engine | Flash v2 (fastest), Flash v2.5 (fast, multilingual), Turbo v2 (balanced), Turbo v2.5 (balanced, multilingual), Multilingual v2 (quality), v3 Conversational | Flash v2 | Full control |
| **Stability** | "Higher = more consistent, lower = more expressive." | 0 to 1, in steps of 0.05 | 0.5 | Instructions |
| **Speed** | "Speaking pace." | 0.7 to 1.2, in steps of 0.05 | 1 | Instructions |
| **Similarity** | "How closely to match the original voice." | 0 to 1, in steps of 0.05 | 0.75 | Instructions |
| **Latency optimisation** (under **Advanced**) | "Higher = faster start, slightly lower quality." | 0 to 4, whole numbers | 3 | Instructions |
| **Audio output format** | The sound format the agent sends | pcm\_8000, pcm\_16000, pcm\_22050, pcm\_24000, pcm\_44100, pcm\_48000, ulaw\_8000 | pcm\_16000 | Full control |
| **Text normalisation** | How numbers and symbols are turned into speech | **System prompt** or **Automatic** | System prompt | Full control |
**The language and the speech engine must match.** An English agent cannot use Flash v2.5 or Turbo v2.5, and an
agent in any other language cannot use Flash v2 or Turbo v2. A save that breaks this rule is refused with a message
that names the fix. Below Full control the message says "Pick a different language, or contact support to change
the voice model for you."
Recommended:
* **Voice:** choose a voice that speaks your callers' language and suits your store. Use **Play sample** to listen
before you choose (see [Choosing a voice](#choosing-a-voice)).
* **TTS model family:** leave it as set. A wrong choice can stop the agent working properly, and it must match the
language. Contact RuleInside if you think it needs changing.
* **Stability:** keep the default unless the voice sounds too flat (lower it for a more expressive voice) or too
changeable (raise it for a more consistent one).
* **Speed:** keep 1 unless callers find the agent too fast or too slow. Change it in small steps and test with
[Preview](/managing-agents#testing-an-agent-with-preview) after saving.
* **Similarity:** keep the default unless the agent does not sound close enough to the voice you chose; then raise
it.
* **Latency optimisation:** keep the default. Higher values give a faster start with slightly lower sound quality.
If the agent feels slow to answer, look at [Turn taking](#turn-taking) first: the editor says that is usually what
people mean.
* **Audio output format:** leave it as set and contact RuleInside if you think it needs changing. A wrong format
breaks the sound on phone calls.
* **Text normalisation:** leave it as set and contact RuleInside if numbers or symbols are read out wrongly.
### Choosing a voice
**Change voice** opens **Select a voice** (use **Back** to return).
* **My Voices**: the voices already in your account. **Explore**: the public voice library.
* Search with "Search your voices…" or "Search the library…".
* Filter with **Language** (English, Italian, Spanish, French, German, Portuguese, Dutch, Polish, Turkish, Arabic,
Hindi, Japanese, Korean, Chinese), **Category** (Professional, Famous, High quality), **Gender** (Female, Male,
Neutral) and **Age** (Young, Middle-aged, Old).
* **Play sample** plays a voice before you choose it, when a sample exists.
* **Explore** shows up to 40 results ("Searching…" while it looks). "No voices found." means nothing matched.
What happens when you pick a voice:
* A voice from **My Voices** only changes the form. It is saved when you click **Save changes**.
* A voice from **Explore** is first added to your voices ("Voice added to your workspace." or, if it fails,
"Could not add that voice.") and then selected. If you already have a voice with the same name, that one is
used. **The voice stays in your voices even if you then discard the change.**
* A voice you just added can take up to 5 minutes to appear under **My Voices** and in the Voice row.
## Language
**Default language** (edit level **Instructions**): "The default language your agent communicates in." Choose from
English, Italian, Spanish, French, German, Portuguese, Dutch, Polish, Swedish, Danish, Norwegian, Finnish, Turkish,
Arabic, Hindi, Japanese, Korean and Chinese. Saves with **Save changes**.
If the agent's current language is not in this list, the row shows its language code. Changing the language must
respect the language and speech engine rule under [Voice](#voice).
Recommended: set **Default language** to the language most of your callers speak. If the save is refused because of
the voice model, contact RuleInside (or, at Full control, change **TTS model family** in the same save).
## Language model
The row shows the model RuleInside set for the agent. You cannot change it. The panel reads "How your agent thinks
and responds."
| Field | What it does | Values | Default | Edit level |
| - | - | - | - | - |
| **Creativity (temperature)** | "Lower = focused and consistent, higher = more varied." | 0 to 1, in steps of 0.05 | 0.5 | Instructions |
| **Max response length (tokens)** | The longest answer the agent may give. Shown only when you can change it. | -1 or 0 for no limit, or 50 to 100,000. 1 to 49 is refused: "Max response length must be -1 (no limit) or at least 50." | -1 (no limit) | Full control |
| **Model** | The model RuleInside set for the agent | Set by RuleInside | | Never |
| **Reasoning** | The model's reasoning setting ("Default" when not set) | Set by RuleInside | | Never |
Recommended:
* **Creativity (temperature):** keep the default. Lower it if you want more focused, consistent answers; raise it
only if answers sound too repetitive.
* **Max response length (tokens):** keep -1 (no limit). If you want shorter answers, ask for them in your
instructions instead (see [Writing instructions](/writing-instructions)).
* **Model** and **Reasoning:** nothing to do. Contact RuleInside if you have a question about them.
## Turn taking
"How long your agent waits before it decides the caller has finished speaking. This is usually what people mean
when they say an agent feels slow." The row summarises the current eagerness and wait, or says "Using default
pacing". All fields need the **Instructions** level.
| Field | What it does | Values | Default |
| - | - | - | - |
| **Response eagerness** | "How readily the agent starts speaking. Eager feels responsive in conversation; Patient gives callers room when they are reading out a number, an address or an email." | Patient - waits longer, Normal, Eager - replies sooner | Normal |
| **Wait for a reply (seconds)** | "If the caller goes quiet for this long, the agent speaks up." | 1 to 30. Cannot be left empty. | 7 |
| **Turn detection model** | How the end of the caller's turn is detected. v3 is the current default and detects the end of a turn more accurately than v2. | v3 - current default, v2 - previous | v3 |
Recommended:
* **Response eagerness:** keep **Normal** for most stores. Choose **Eager** if the agent feels slow to answer.
Choose **Patient** if callers often read out order numbers, addresses or email addresses and the agent starts
talking before they finish.
* **Wait for a reply (seconds):** keep 7. A shorter wait makes the agent speak up sooner when the caller goes quiet;
a longer one gives callers time to find something, such as an order number.
* **Turn detection model:** keep v3.
## Call limits
"How long a call may last and when the agent hangs up. Longer calls use more minutes." The row shows the
maximum, for example "Up to 10 min per call (default)". Every field you can change needs the **Instructions**
level.
| Field | What it does | Values | Default |
| - | - | - | - |
| **Maximum call length (minutes)** | "The call ends when it reaches this length." | 1 to 120. Cannot be left empty. | 10 |
| **What the agent says when time is up** | The goodbye said when the limit is reached | Up to 500 characters. "Optional. Leave it empty to end the call without a message." | Empty |
| **Hang up when the caller stays silent** | A switch that ends the call after a silence | On or off | Off |
| **Seconds of silence** | "Ends the call after this many seconds without the caller speaking." | 10 to 120 | 30 when you switch it on |
| **Wait at the start of the call (seconds)** | "Only used when the caller speaks first: how long the agent waits for them to speak." | 1 to 30, or empty to use the reply wait from Turn taking | Not set |
| **When a call reaches voicemail** | What happens on voicemail | Display only | |
Recommended:
* **Maximum call length (minutes):** keep 10 for most stores. Raise it if real conversations reach the limit (check
the **Duration** column on [Calls](/conversations)). Lower it if you want to cap the minutes a single call can use.
* **What the agent says when time is up:** write a short, polite goodbye, so the call does not end without a word.
* **Hang up when the caller stays silent:** consider switching it on if your calls show long stretches where the
caller has stopped talking.
* **Seconds of silence:** start with 30. We suggest keeping it longer than **Wait for a reply (seconds)** in
[Turn taking](#turn-taking), so the agent speaks up before the call is ended.
* **Wait at the start of the call (seconds):** leave it empty unless you use **Caller speaks first** and want a
different wait at the start.
* **When a call reaches voicemail:** nothing to change here. Contact RuleInside if you need it changed.
**When a call reaches voicemail** can never be changed here. It shows one of "Voicemail detection is off for this
agent.", "The agent hangs up without leaving a message." or a message the agent leaves, with "Contact RuleInside to
change it." It follows the **Voicemail detection** switch on the [Tools](/tools) page. RuleInside's policy is that
the agent hangs up and leaves no message.
## Conversation behaviour
"Interruptions, filler while thinking, background sound, words to recognise and keypad input." The row shows
"Interruptions on" or "Interruptions off". Every field needs the **Instructions** level. Where a field was never
set, the default shown is the value the call actually uses.
### Interruptions
| Field | What it does | Values | Default |
| - | - | - | - |
| **Callers can interrupt the agent** | "When on, the agent stops talking as soon as the caller speaks." | On or off | On |
| **Callers cannot interrupt the greeting** | "The first message is always played to the end, for example so the recording notice is heard." | On or off | Off |
| **Words that do not interrupt** | "Short replies such as "ok" or "got it" that should not stop the agent." | One per line, up to 50 lines, each up to 50 characters. Blank lines are dropped. | Empty |
| **Also use the standard list of such words** | Adds a standard list of such words to yours | On or off | Off |
| **Extra patience while spelling** | "Gives callers more time when they spell an order number, a name or an email address." | **Automatic** or **Off** | Automatic |
If the current interruption setting could not be read, the switch is locked and says "The current setting could
not be read, so it cannot be changed right now." The row then shows "Settings".
Recommended:
* **Callers can interrupt the agent:** keep it on, so the agent stops talking as soon as the caller speaks, as a
person would on the phone.
* **Callers cannot interrupt the greeting:** switch it on if your greeting carries the recording notice and you
want it always heard in full.
* **Words that do not interrupt:** add the short replies your callers use, in their language, if the agent stops
talking when they say "ok" or "yes".
* **Also use the standard list of such words:** switch it on as a quick start, then add your own words if needed.
* **Extra patience while spelling:** keep **Automatic**.
### Filler while thinking
| Field | What it does | Values | Default |
| - | - | - | - |
| **Say a short filler when an answer takes a moment** | "Avoids dead air on slow answers." | On or off | Off |
| **Say it after (seconds)** | How long to wait before the filler | 0.5 to 8, in steps of 0.5 ("3 works well") | 3 when you switch it on |
| **Filler** | What the agent says (shown while the switch is on) | Up to 200 characters. Required while on. | "Hhmmmm...yeah." |
Recommended: switch the filler on if callers hear silence while the agent works out an answer, for example while
it looks up an order. Keep **Say it after (seconds)** at 3 (the portal says "3 works well"), and write a short
phrase in the agent's language, such as "One moment...".
### Background sound
| Field | What it does | Values | Default |
| - | - | - | - |
| **Background sound** | A sound played behind the agent's voice | None, Office 1, Office 2, Restaurant, City, Typing, Elevator music 1, Elevator music 2, Elevator music 3, Elevator music 4 | None |
| **Background volume** | Its volume (shown only when a sound is chosen) | 0.01 to 1 | 0.15 |
Recommended: keep **None** unless you want a sound behind the agent's voice. If you choose one, keep the volume low,
close to the default, so callers can hear the agent clearly.
### Recognition
**Words to recognise**: "Brand and product names the agent should catch reliably." One per line, up to 50 lines,
each up to 50 characters.
Recommended: add your store's name and the product or brand names callers say most often, especially ones with an
unusual spelling.
### Keypad (phone calls)
| Field | What it does | Values | Default |
| - | - | - | - |
| **Wait for the next key (seconds)** | "How long the agent waits for more digits." | 0.5 to 10, in steps of 0.5 | 2 |
| **Pressing # finishes the input** | Ends the caller's keypad entry when they press # | On or off | On |
| **Hide typed digits in transcripts** | "Replaces the digits a caller types with a marker in transcripts and call logs." | On or off | Off |
Recommended:
* **Wait for the next key (seconds):** keep 2. Raise it if callers type long numbers slowly and get cut off.
* **Pressing # finishes the input:** keep it on.
* **Hide typed digits in transcripts:** switch it on if callers type numbers on the keypad that you do not want in
your transcripts and call logs.
**Hide typed digits in transcripts** hides only the digits the caller types. Digits the agent reads back aloud,
or passes on to a tool, are not hidden.
Saving one keypad, filler or background field never resets the others in the same group.
## Knowledge base and Tools cards
These two cards in the right column are summaries.
* **Knowledge base** ("Documents your agent answers from.") lists the documents attached to the agent, or "No
documents attached yet." It is read-only and has no link. To add or change documents, open
[Knowledge](/knowledge) in the sidebar.
* **Tools** lists every tool on the agent with a type badge: **Store**, **In-page**, **Built-in**,
**Integration** or **Tool**. On a Full control agent it reads "What this agent can do during a call." and
**Manage tools** opens the [Tools](/tools) page. Below Full control it reads "What your agent can do, managed by
RuleInside." The tools shown here as **Store** are the ones labelled **Webhook** on the Tools page.
## Common questions
Panels do not save when you close them. Click **Save changes** in the bar at the bottom of the page.
A field in one of the panels is empty or out of range. The most common cause is an empty **Wait for a reply
(seconds)** or **Maximum call length (minutes)**. Open the highlighted panel to see the message under the
field.
Either you are a member rather than an owner, or the agent's edit level does not include that setting. See
[Edit levels](#edit-levels). Your main agent is often below Full control.
They need **Full control**, because a wrong value can break the sound on phone calls. On your main agent,
RuleInside usually keeps them. Contact RuleInside if you need them changed.
The current speech engine does not support that language. At Full control, change **TTS model family** in the
same save. Below Full control, contact RuleInside to change it for you.
Open **Turn taking**, set **Response eagerness** to Eager and try a shorter **Wait for a reply (seconds)**.
Add those words to **Words that do not interrupt** in **Conversation behaviour**, or switch on **Also use the
standard list of such words**.
Check **Maximum call length (minutes)** and **Hang up when the caller stays silent** in **Call limits**.
Yes. Under **Recording notice**, choose **After the first exchanges** and write the sentence in **What the agent
says**. This needs the **Instructions** level. The greeting must still say the caller is talking to an AI
assistant.
Yes. Under **Who speaks first**, choose **Caller speaks first**. You can also set how long it waits in **Wait
for the caller to speak (seconds)**. This needs the **Instructions** level.
On phone calls, when RuleInside has switched it on for your account. You can add up to 5 mobile, landline or
toll-free numbers in the same country as your store's phone number.
Transfer to a person is not switched on for your account. Removing a transfer is always allowed, so you can stop
one without contacting us. To add or change one, contact RuleInside.
The storefront did not pass in a value for that variable, so its default was used. Check the default under
**Personalization variables**.
On the [Knowledge](/knowledge) and [Tools](/tools) pages, not in the editor.
# API keys
Source: https://docs.ruleinside.com/api-keys
Create keys so your own systems can talk to RuleInside, each with only the permissions it needs.
An API key is a long password that lets your own software talk to RuleInside. Each key carries only the
permissions you choose when you create it, so you can give each integration its own key and revoke it on its own.
Open **Settings**, choose the **Developer** tab, and find the **API keys** card. The same tab holds
[Webhooks](/webhooks).
## Who can do what
* **Everyone on your team** can see the list of keys.
* **Only the owner** can create or revoke a key. Members see "Only an account owner can create or revoke API keys."
If your workspace is not linked to a store yet, API keys cannot be created.
## The list of keys
The list shows every key that has not been revoked, including the [licence keys](/plugin) of the WordPress plugin.
Each row shows:
* the start of the key followed by "…", for example `ri_sk_live_xxxxxx…` (the full key is never shown again);
* one badge for each permission the key has;
* **Created** and the date;
* **Last used** with the date and time, or **Never used**;
* **Revoke** (owner only).
Revoked keys are hidden; a line below the list says how many. With no keys you see "No API keys yet."
If you see "We could not load your API keys", the list could not be loaded at that moment. Your existing keys are
unaffected and still work. Try again shortly rather than creating a replacement.
## Create an API key
Owner only. The **Create an API key** dialog opens.
Choose only the permissions this integration needs (see the table below). **Create key** is available once at
least one is ticked.
It reads "Creating…" while the key is made. If it fails, you see "Could not create the key."
The **Copy your API key now** dialog shows the full key, starting with `ri_sk_live_`. Click **Copy** (it
changes to **Copied**), store the key safely, then click **Done**.
The key is shown only once. We store only a scrambled fingerprint of it, so nobody, including us, can show it to
you again. If you lose it, create a new key and revoke the old one.
## The permissions
Every permission can read or change your customer data, so grant only what the integration actually needs.
Permissions marked **Can place calls** can also start calls, which costs money and contacts your customers.
| Permission | Label in the portal | What it allows |
| - | - | - |
| `contacts:read` | **Read your contacts** | Read your contact records. |
| `contacts:write` | **Create and update contacts** | Create contacts and change them. |
| `consent:write` | **Record consent and opt-outs** | Record consent and opt-outs. This also creates or updates the contact. |
| `calls:read` | **Read call history and transcripts** | Read your calls, including customer details and summaries. |
| `calls:trigger` | **Start outbound calls** (**Can place calls**) | Start a call, and launch a campaign. |
| `campaigns:read` | **Read campaigns** | Read your campaigns, including their test phone numbers. |
| `campaigns:write` | **Create and launch campaigns** | Create campaigns as drafts and cancel them. Launching a campaign also needs **Start outbound calls**. |
| `webhooks:manage` | **Manage webhook endpoints** | Add and remove [webhook endpoints](/webhooks) through the API. An endpoint decides where your customer data is sent. |
| `events:write` | **Send events into RuleInside** (**Can place calls**) | Send events. An event can schedule a follow-up call if you have an [Automation](/automations) that matches it. |
The plugin download permission (`plugin:download`) is not offered here. Create it as a licence key on the
[Plugin](/plugin) tab instead.
You cannot change the permissions of an existing key. Create a new key with the permissions you want, move your
integration to it, then revoke the old key.
## Using a key
Send the key in the `Authorization` header of every request:
```http Header theme={null}
Authorization: Bearer ri_sk_live_xxxxxx
```
* Keys created here are always live keys, starting with `ri_sk_live_`.
* A key has no expiry date. It stops working when you revoke it, or while your account is not active.
* Each key can make a set number of requests a minute, depending on your plan. Above that, the API answers `429`
with a `Retry-After` time to wait before trying again.
| Plan | Requests a minute, per key |
| - | - |
| Lite | 60 |
| Starter | 120 |
| Growth | 300 |
| Scale | 600 |
| Pay-as-you-go | 120 |
## Revoke a key
Click **Revoke** on the key and confirm with **Revoke key** ("Revoke this API key?"). You see "API key revoked.",
or "That key was already revoked, or no longer exists."
Anything using the key stops working immediately, and this cannot be undone. Revoking a key with the
`plugin:download` badge also stops plugin updates for any store using it.
## Common questions
Only the ones it uses. Anything that can start calls is marked **Can place calls**: grant those only to an
integration that must place calls.
It cannot be shown again. Create a new key, switch your integration over to it, then revoke the old one.
Check whether the key was revoked (the line below the list counts revoked keys), whether your account is active,
and whether the integration is over its requests a minute (it then gets `429`).
They are licence keys for the WordPress plugin. Revoking one stops plugin updates for the store that uses it.
See [WordPress plugin](/plugin).
# Automations
Source: https://docs.ruleinside.com/automations
Rules that place calls for you when contacts match a condition, with no campaign to launch.
**Automations** are standing rules that place calls on their own when contacts match a condition: "call if there's
been no activity for a while", "call a few days after an event", or "call contacts matching a field". There is
nothing to launch: a rule that is switched on runs by itself.
Every call a rule places passes the same checks as any other call. Nobody is called who opted out, is on your
do-not-call list, has no recorded consent, is outside their calling hours or days, or was called too often recently,
and nobody at all is called once your daily call limit is reached or while your store has not confirmed its calling
consent. Switching a rule on never grants consent. See [How calling rules protect your customers](/outbound-rules).
## Who can do what
* **Any member** can see the rules and use **Preview**, which never places a call.
* **Only the account owner** can add a rule, switch it on or off, or delete it. Other members see **Only account
owners can add or change automations: ask an owner of your team.** and a read-only **Active** or **Off** badge.
## The rules on the page
The line at the top counts your rules (for example **3 automations.**, or **No automations yet.**). Each rule is a
card, newest first, with:
* its trigger: **No activity for a while**, **Follow up after an event**, **Contacts matching a field** or **Win back
past customers**;
* its status: **Active** or **Off**;
* a one-line description, for example **Call anyone with no activity for 30 days.**;
* (owner) a switch, **On** / **Off**;
* **Preview** (**Checking…** while working);
* (owner) **Delete**.
## Creating a rule
Click **New automation** (owner only). **Set a rule in plain terms. Nothing is called until it's switched on: and
every call still passes the consent & calling-hours check.**
### What should trigger a call?
Default: **No activity for a while**. Each trigger has its own fields:
| Trigger | Field | Default | Allowed values |
| - | - | - | - |
| **No activity for a while** | **Call if there's been no activity for (days)** | 30 | A whole number, 1 or more. |
| **Follow up after an event** | **Event type** (for example `appointment.booked` or `order.completed`) | empty | Required. Letters, digits, dots, underscores and hyphens, starting with a letter or digit, up to 64 characters. It must match the event's name exactly. |
| | **Days after the event** | 3 | 0 to 365. |
| **Contacts matching a field** | **Field** (a contact field, for example `deal_stage`) | empty | Required. A valid contact column name. |
| | **Equals** (for example `qualified`) | empty | Required, not blank. The value must match exactly. |
| **Win back past customers** | **Last ordered more than (days) ago** | 60 | A whole number, 1 or more. |
What each trigger looks at:
* **No activity for a while** uses the last time an event about the contact arrived. Only contacts that were once
active and then went quiet qualify.
* **Win back past customers** uses the date of the contact's last order in your store. Only contacts with a recorded
store order qualify. You can have only one win-back rule switched on at a time.
* **Contacts matching a field** uses the contact's own columns (see [Contacts](/contacts)).
* **Follow up after an event** calls a contact the chosen number of days after an event of that exact type arrives
for them (see [How rules run](#how-rules-run)).
### Which agent calls?
**Default agent** (your default agent at the moment of each call), or one of your agents by name. Choosing the agent
marked **(default)** by name keeps that agent even if you change your default later. The call always comes from the
number assigned to that agent; there is no number to choose.
### Switch on now
**Switch on now** is off by default: the rule is created switched off, and you switch it on from its card when you
are ready.
### Saving
* **Cancel** closes the form. Nothing is saved.
* **Create automation** (**Creating…** while working) saves the rule. If a required value is missing you see **Fill
in the rule details first.**; on success, **Automation created.**
Each run of a rule calls at most 2,000 people, and uses the default retry settings for unanswered calls (up to 3
attempts, at least 3 hours apart, at most 2 a day). These cannot be changed in the portal.
## Preview
**Preview** shows who a rule would reach right now. It never places a call, and any member can use it.
* For **No activity for a while**, **Contacts matching a field** and **Win back past customers**: **\{matched}
contact(s) match right now · \{callable} callable after the compliance check. (Preview places no calls.)**, then
**Excluded:** with each reason and its count. If the group was cut at the rule's limit: **Showing a capped sample:
the real group may be larger.**
* For **Follow up after an event**: **This runs when an "\{eventType}" event arrives: it then calls that contact
\{days} day(s) later (if they're reachable). No one is called right now.**
Preview runs every check as of now, including calling hours, so the callable number can change with the time of day.
The reasons after **Excluded:** are short codes:
| Shown | Meaning | What to do |
| - | - | - |
| no store attestation | Your store has not confirmed its calling consent. | The owner confirms it under **Settings › Calling consent**. |
| opt out | The person opted out. | Nothing. |
| dnc | The number is on your do-not-call list. | Nothing. |
| no consent | No recorded lawful basis to call. | Record consent in [Contacts](/contacts). |
| outside window | Outside their calling hours or days now. | Nothing: it changes with the clock. |
| no timezone | Their time zone could not be worked out. | Store a time zone, or set a default under **Settings › Calling consent**. |
| frequency cap | Called too often in the last 7 days. | Nothing: it clears within a week. |
| daily cap | Your daily call limit is reached. | Nothing: it clears within 24 hours. |
| error | A check could not be completed. | Try again shortly. |
## Switching a rule on or off
The owner's **On** / **Off** switch saves at once, with no confirmation. If the save fails, the switch flips back and
you see **Could not update the automation.** Switching a rule off stops its future runs; it does not recall calls
already placed.
## Deleting a rule
**Delete** asks **Delete this automation?** "It will stop running. This can't be undone." Press **Delete** to confirm,
or **Keep it** to go back. Calls already placed are not recalled.
A rule cannot be edited. To change one, create a new rule with the settings you want, then delete the old one.
## How rules run
* **No activity for a while**, **Contacts matching a field** and **Win back past customers** run twice a month: at
04:00 UTC on the 1st and on the 15th. A rule you switch on in between waits for the next run.
* **Only one of these rules runs for your store each time: the oldest one that is switched on.** A newer rule
switched on at the same time does nothing while the older one is on. Keep one of these rules on at a time, and
switch the others off.
* A run is skipped when your previous automatic batch is still running, or when the rule's agent has no outbound
number assigned. Assign one on [Phone numbers](/phone-numbers).
* Rules use your **account's** calling hours and days (under **Settings › Calling consent**). A rule has no hours of its
own.
* **Follow up after an event** does not wait for those dates. When an event of that exact type arrives for a
contact, through a [Connection](/connections) or the developer API, one call is planned for the event's time plus
the days you chose, and placed after every check runs. An event whose call would already be more than 2 days late
(for example an old event sent late) is ignored.
* Orders from your store count as events only when store activity events are switched on for your account by our
team (they are off by default). Otherwise an `order.completed` rule only fires for events sent through a
Connection or the developer API.
* An event follow-up created in the portal is always a phone call.
## Common questions
Scheduled rules run only at 04:00 UTC on the 1st and 15th of each month. Also check that it is the oldest rule
switched on (only that one runs), that your previous automatic batch has finished, and that the rule's agent has
an outbound number assigned.
The **Event type** must match the event's name exactly, for example `order.completed`, and events must actually
arrive: through a Connection or the developer API, or from your store's orders only if store activity events are
switched on for your account. An event more than 2 days past its due time is ignored.
Read **Excluded:**. "no store attestation" is fixed under **Settings › Calling consent**, "no consent" in
[Contacts](/contacts), "frequency cap" clears by itself within a week, "outside window" changes with the time of
day, and "error" means try again.
No. Create a new rule with the settings you want, then delete the old one.
Only one win-back rule can be switched on at a time. Switch the other off, or delete it, before switching a new
one on.
No. Rules always use your account's calling hours and days, in each person's own time zone. To call at other
times, use a [campaign](/campaigns) with its own calling hours.
**No activity for a while** looks at the last event about the contact. **Win back past customers** looks at the
date of their last order in your store. They can select different people.
At most 2,000 per run, and only those who pass every check at that moment.
# Plans & billing
Source: https://docs.ruleinside.com/billing
The plans, billing interval, monthly minutes, agents to build, paying, invoices, managing or cancelling your plan, and what happens at a limit.
RuleInside is a **managed service**. Every plan comes with conversation minutes, and our team builds and manages
your agents for you. You choose and manage your plan on the **Billing** page, under the **Settings** heading in the
left sidebar.
For current prices and exactly what each plan includes, see [ruleinside.com](https://ruleinside.com) or ask our
team.
## Who can manage billing
* **Owners** (any owner, not only the primary owner) can choose a plan, see invoices, open the billing portal and
cancel.
* **Members** cannot. Before your organization has a plan, they see **You cannot manage billing** instead of the
plans. Once you have a plan, they see only the **Your Plan** card and the **Minutes** card. See
[Who can do what](/roles).
## The plans
Managed plans, billed **monthly** or **yearly**. Each includes a block of minutes every month; usage beyond it
is billed per minute.
**Minute bundles** in ten sizes, billed **monthly**. When a bundle runs out it is bought again automatically,
at most twice in a billing period.
| Plan | Billing interval | Minutes included each month |
| - | - | - |
| Pay-as-you-go bundle | Monthly | 330, 500, 750, 1,100, 1,500, 2,000, 2,600, 3,200, 4,000 or 5,000, as you choose |
| Lite | Monthly or yearly | 1,000 |
| Starter | Monthly or yearly | 2,000 |
| Growth (marked **Most Popular**) | Monthly or yearly | 4,400 |
| Scale | Monthly or yearly | 6,400 |
No current plan has a free trial. Some organizations also see a plan arranged specially for them; it appears only
to the organization it was made for.
There is no charge per member, and no limit on the number of agents. See [Who can do what](/roles) and
[Plans and limits](/limits).
## Choosing a plan
Until your organization has a plan, the **Billing** page shows the card **Manage your Organization Plan** with a
card for each plan. Once you subscribe, the plan picker disappears.
### Billing interval
At the top, switch between **Billed monthly** and **Billed yearly**. Only the intervals that exist are shown:
pay-as-you-go bundles appear on the monthly tab only. On the yearly tab, each card shows its price as a monthly
equivalent, with the yearly total underneath.
If the cards do not fit on your screen, use the arrows (**Show previous plans** and **Show more plans**) to scroll.
### What each plan card shows
* A round selector and the plan name. Clicking the name also selects the card.
* A short description and the headline price.
* A list of what is included.
* **Details**: the usage lines, including the minutes included in the plan and the rate for each minute above
them.
* The **Proceed to Payment** button.
### Monthly minutes (pay-as-you-go only)
The pay-as-you-go card has a **Monthly minutes** dropdown with the ten bundle sizes. **It starts at the smallest,
330 minutes.** As you change it, the card updates three lines:
* the minutes a month, for example "330 minutes a month";
* **About ... calls**: the minutes divided by 3, rounded down to the nearest 10. For 330 minutes, "About 110
calls";
* **Or ... message conversations, same pool**: the minutes times 2. For 330 minutes, "Or 660 message
conversations, same pool".
The calls and message conversations come out of the same minutes; they are not extra.
### AI agents to build (managed plans only)
The Lite, Starter, Growth and Scale cards have an **AI agents to build** dropdown: how many agents our team builds
for you when you subscribe.
| Option | Meaning |
| - | - |
| **None, subscription only** | No build fee. Choose this if we already run agents for you, or to add one later. |
| **1 agent** (the default) | One agent built for you. |
| **2 agents** to **10 agents** | That many agents built for you. |
The build fee is charged **once**, on your first invoice, and the card shows the total under the dropdown. The
number is **not a limit**: you can have more agents later. An extra agent is billed, never refused.
## Paying
Pick the billing interval, select a plan card, and set **Monthly minutes** or **AI agents to build**.
The button shows **Redirecting…** and the other cards lock, so only one payment can start.
A secure payment form opens inside the page. On managed plans you can enter a promotion code there.
Pay-as-you-go bundles do not take promotion codes.
You see **Done! You're all set.** and a confirmation email is on its way. Click **Proceed to App** to return to
**Billing**.
After your first subscription, our team sets up your agent. Until it is ready, your dashboard shows that it is
still being set up.
If something goes wrong, you see "Please choose a plan" (no card was selected) or **Error requesting checkout**.
Try again a little later.
## After subscribing
### Your Plan
Everyone in the organization sees the **Your Plan** card: the plan name, a status badge, the description and the
**Details** lines. It also shows a notice whenever your status is not **Active**.
| Badge | What it means | Your assistant |
| - | - | - |
| **Active** | Your subscription is active. | Keeps working |
| **Trial** | You are on a trial. | Keeps working |
| **Past Due** | Your invoice is past due. Update your payment method. | Keeps working while the payment is retried |
| **Unpaid** | Your invoice is unpaid. Update your payment method. | Paused |
| **Canceled** | Your subscription is canceled. | Paused |
| **Incomplete** | We are waiting for your payment to go through. | Paused |
| **Expired** | Your payment expired. Update your payment method. | Paused |
| **Paused** | Your subscription is paused. | Paused |
When you have scheduled a cancellation, the card also shows **Subscription Cancelled** with the date your plan
ends. What a paused account can and cannot do is on [Plans and limits](/limits#when-a-payment-fails-or-your-subscription-ends).
### Invoices
Owners see the **Invoices** card once your first payment is set up.
* **Upcoming invoice** shows the date and amount of your next invoice, when there is one.
* The table lists your most recent invoices, up to 24, with **Date**, **Amount**, **Status** and **Invoice**.
Click the download icon in **Invoice** to open the PDF in a new tab.
* The status is one of **Paid**, **Open**, **Void**, **Uncollectible** or **Draft**.
* For older invoices, use the billing portal (below).
If the card shows **No invoices yet** when you expected some, reload the page a little later. The list is fetched
from our payment provider each time, and the card shows empty if it could not be reached.
### Manage your Billing Details
Owners can click **Visit Billing Portal**. Our payment provider's billing portal opens in a new tab (or in the same
tab if your browser blocks pop-ups). There you update your card and billing address, download invoices and change
your plan.
## Changing your plan
Once you have a plan, the plan picker is no longer on the **Billing** page. To change plan, use **Visit Billing
Portal**, or contact us.
## Cancelling your plan
Owners see the **Cancel your plan** card while the plan is **Active**, **Trial**, **Past Due** or **Unpaid**.
The dialog **Cancel your plan?** shows the date your plan stays active until.
Click **Cancel at the end of my term**, or **Keep my plan** to change your mind.
You see "Your plan is scheduled to end. You keep full access until then."
* You are not charged again. Your plan keeps working until the end of the term you have already paid for. On a
yearly plan, that is the end of the year.
* Usage beyond your included minutes is still billed as normal until the end date.
* **Minutes you have not used end with your plan. They are not refunded.**
* Until the end date, you can undo it: the card becomes **Keep your plan**, with the button **Don't cancel**. You
then see "Your plan will continue. The scheduled cancellation has been removed."
Cancel with the **Cancel plan** button on this page rather than in the billing portal. It always ends your plan at
the end of the longest term you have paid for, so a yearly plan keeps its full year.
If you reload the page straight after cancelling, it may briefly show **Cancel plan** again while the change is
confirmed. Wait a moment and reload before trying again. If you see "We could not confirm your cancellation",
reload the page to check whether it went through before trying again.
## Minutes and what happens at a limit
### Your minute balance
The **Minutes** card on the **Billing** page shows the call minutes on your account. Everyone in the organization
sees it.
* **... minutes left**: the whole minutes you have left, rounded down.
* Under it, when the next expiry is known: "... of them expire on \[date] if they are not used before then.", or
"None of them are due to expire."
Each period's minutes last twelve months, and the ones closest to expiring are used first. If your plan ends, any
unused minutes end with it. Calls from the last day may not be counted yet.
If the card is not shown, your workspace is not linked to a store yet, or your balance could not be read at that
moment: reload the page later, or contact our support team (the **Support** link at the top of this site).
### Managed plans
Your plan includes a block of minutes each month, on yearly plans too. When you use more, the extra minutes are
billed per minute, at the rate shown in **Details**. Your assistant keeps working: nothing is paused because the
included minutes ran out.
### Pay-as-you-go bundles
When your bundle runs out, the same bundle is bought again with your saved card. This happens at most twice in one
billing period.
* After the second extra bundle in a period, outbound calls and messages can be stopped for the rest of that
period. Inbound calls keep working.
* If your card is declined, everything can be stopped, inbound calls included, until the payment is sorted out.
* If your bank asks you to confirm a payment, you have 24 hours to do it.
The full detail, including the notices you receive at 80% and 100% of a bundle, is on
[Plans and limits](/limits).
### Limits on every plan
Each plan also sets how many outbound calls you can place in a day and how many can run at the same time. See
[Plans and limits](/limits#daily-calls-for-your-account).
## Common questions
Look at the **Minutes** card on the **Billing** page. It also shows how many of them expire next, and when. Calls
from the last day may not be counted yet. See [Your minute balance](#your-minute-balance).
Once you have a plan, use **Visit Billing Portal** on the **Billing** page, or contact us.
Click **Cancel plan** on the **Billing** page. You keep full access until the end of the term you paid for, and
you can undo it until then with **Don't cancel**.
Not on any current plan.
Yes on Lite, Starter, Growth and Scale: enter it in the payment form. Pay-as-you-go bundles do not take
discount codes.
No. It is how many agents our team builds for you at the start, charged once. You can add more agents later.
They are a member. Billing is for owners. Make them an owner on the [Members](/members) page, or download the
invoices yourself.
On a pay-as-you-go bundle, the most likely reason is a declined card. Update your card with **Visit Billing
Portal** and contact us.
No. Only cancelling the plan stops billing. See [Cancelling your plan](#cancelling-your-plan).
No. Minutes you have not used end with your plan, and they are not refunded.
Need a custom arrangement, for example very high volume? Get in touch with our team (the **Support** link at the
top of this site) and we will tailor a plan.
# Calling consent and calling hours
Source: https://docs.ruleinside.com/calling-hours
Confirm your store may have an AI agent contact its customers, and set when your agent may call.
Open **Settings** and choose the **Calling consent** tab. The page has two cards:
* **Calling consent**, where an owner confirms, for the whole account, that the store has a lawful basis to contact
the people it asks RuleInside to contact. Until this is confirmed, **nothing outbound is sent**.
* **Calling hours and days**, which sets when your agent may call your customers.
If your workspace has no store yet, the page shows "This workspace is not linked to an account yet." and nothing else.
## Who can change this page
Everyone in your workspace can see this page. Only an **owner** can confirm, withdraw, or change calling hours and
days. Other members see the controls greyed out and "Only an owner of this workspace can change this." See
[Who can do what](/roles).
## The Calling consent card
The card explains: "Before this account can place automated calls, you confirm you have a lawful basis for calling
each person on your list. That can be consent they gave you, or a list you obtained lawfully from someone else. It
has to cover automated AI voice calls where you trade."
This confirmation is separate from the consent you record on each contact, and **you need both**:
* the consent on a contact says on what basis that person may be contacted (see [Contacts](/contacts));
* this confirmation is you, as the owner, taking responsibility for the whole list.
Until an owner confirms, **no calls and no WhatsApp messages are sent**, even to contacts marked **Calls:
callable**. Calls and messages held back for this reason are retried a limited number of times rather than cancelled
at once, but do not rely on that: confirm before you launch a campaign.
### Status
| What you see | What it means |
| - | - |
| **Confirmed**, with "Confirmed for" the country, and the date and time | Active. Outbound calls and messages can run, subject to the checks on each contact. |
| **Not confirmed**, with a red "Outbound calling is paused" alert | Nothing is sent. The alert gives one of the reasons below. |
| "Please re-confirm" | We have updated the wording since you confirmed. **Your calls are still running.** Read the new statement and confirm again when you can. |
| "We could not load this" | "This is a problem on our side, not a sign that anything is wrong with your account. Please reload in a moment." |
**Why outbound calling can be paused:**
| Reason | What the alert says | What to do |
| - | - | - |
| Never confirmed | "You have not confirmed this yet. No calls or messages will be sent until you do." | Confirm. |
| Withdrawn | "You withdrew this confirmation. Nothing will be sent until you confirm again." | Confirm again when you are ready. |
| Country changed | "You confirmed this for" one country, "but the account is now set to" another. "Consent rules differ by country, so please confirm again" for the new one. | Confirm for the new country. |
| No country | "Your account has no country set, so we cannot tell which rules apply. Choose one below and confirm." | Choose the country and confirm. |
| Could not check | "We could not check your confirmation, so nothing is being sent. This is on us. Please try again shortly." | Try again shortly. |
Your dashboard shows the same warning. When paused: "Your account cannot place automated calls until you confirm you
have a lawful basis for calling the people you want us to contact. Nothing is being sent in the meantime." When the
wording has changed: "We have updated the wording of the calling-consent confirmation. Your calls are still running.
Please read it and confirm again when you have a moment." Both have a **Review and confirm** button that opens this
page.
### Fields
| Field | What it does | Values and default |
| - | - | - |
| **Country your account trades in** | The country whose consent rules apply. "This decides which consent rules apply. Change it here if that ever changes." | A searchable list. Type the country's name (with or without accents), another common name for it, or its two-letter code. If nothing matches: "No country matches that. Try the country's own name, or its two-letter code." Starts as "Select a country", or shows the country you confirmed, or the country already set on your account. |
| The statement | The exact text you are confirming, for the country you picked. Before you pick one: "Choose a country to read the statement you are confirming." | Written by RuleInside, naming your country. Stored word for word when you confirm. |
| **I have read the above and confirm it is true.** | Required tick box. | Cleared after each successful confirmation. |
"I confirm that, for every person I ask RuleInside to contact, I have a lawful basis under the law of Italy (IT) to
contact them by telephone with an automated AI voice agent, and that these contacts were obtained in a way that
permits such calls. That basis may be consent given to me directly, or consent obtained by someone else and
lawfully passed to me, for example through a purchased or rented marketing list, provided it covers a call of this
kind being made for my account.
I confirm that establishing that basis, keeping the record of it, and judging whether it is valid in Italy (IT)
are my responsibility as the account owner, not RuleInside's. I understand that RuleInside applies do-not-call,
opt-out and calling-time suppression on my behalf, but does not verify where my contacts came from or whether I
may lawfully call these people.
If a person or an authority ever raises a question or a complaint about a call made for my account, it is mine to
answer and to resolve, including any claim, fine or cost arising from my decision to contact that person, and
RuleInside will provide the call and consent records it holds to help me do that."
RuleInside does not give legal advice. The statement records what you confirm. Whether your basis is valid in your
country is for you to judge.
### Confirming
Pick it under **Country your account trades in**. The statement for that country appears.
Read it in full. It is what you are confirming.
Tick **I have read the above and confirm it is true.**
The button reads **Confirm**, or **Confirm again** if you have confirmed before. It is greyed out until a country
is chosen and the box is ticked. You see "Confirmed. Outbound calling is cleared for this account."
If saving fails you see "Could not save that. Please try again." or a more specific message, for example: "Only an
owner of the organisation that owns this store can confirm this. Ask an owner to sign it." If you see "This store is
linked to more than one organisation, so it is ambiguous who may sign for it." or "This store is not linked to an
organisation, so no one can sign for it.", contact support.
**Confirming also sets your account's country** to the one you chose. If you start trading in a different country,
choose it and confirm again.
### Withdrawing
**Withdraw** appears only when the confirmation is active. It opens a second step titled "Read this before you
withdraw", with the withdrawal statement for the country on record:
> I withdraw the calling-consent confirmation previously given for Italy (IT). From this point my account no longer
> confirms that it holds a lawful basis for contacting people by an automated AI voice agent. I understand that no
> new outbound calls or messages will be started for this account until a new confirmation is given, and that
> anything already scheduled may still be delivered.
"This is recorded exactly as written, and cannot be edited afterwards."
* **Yes, withdraw** (red) withdraws the confirmation. You see "Withdrawn. Nothing further will be sent until you
confirm again."
* **Cancel** goes back without changing anything.
The withdrawal is for the country on record, not for whatever is selected in the country list at that moment.
After you withdraw, no new calls or messages are started, but **anything already scheduled may still be
delivered**, as the statement says.
### What is kept on record
Every confirmation and every withdrawal is kept as a new, permanent record: what was done, the country, the exact
statement, who signed and when. Nothing earlier is edited or deleted. The newest record decides whether outbound
calling is active.
## The Calling hours and days card
"When your AI agent may call your customers. Each customer is called in their own local time."
Calling hours and days decide when **phone calls** are placed. Two kinds of WhatsApp message also wait for them: a
**Marketing** template, and the WhatsApp message a campaign sends after a call. Other WhatsApp messages are not held.
A Marketing follow-up in a conversation the customer started goes at once only within 24 hours of that
conversation; after that, it waits too. See [WhatsApp](/whatsapp).
| Field | What it does | Values and default |
| - | - | - |
| **Account calling hours** | Heading for the hours. | |
| **Set my own calling hours** | Unticked: "Using the standard hours: 08:00 to 21:00, in each customer's local time." Ticked: two hour pickers, a start "to" an end. | **Default: unticked**, so 08:00 to 21:00. Start from 00:00 to 23:00, end from 01:00 to 24:00, whole hours only. The start and end cannot be the same hour. Untick and save to return to the standard hours. |
| **Account calling days** | Seven switches, Mon to Sun. | **Default: every day.** At least one day must stay on. |
| **Timezone for customers we cannot place** | The time zone used only for customers whose own time zone cannot be worked out from their number or profile. | **Default: "None: skip these customers".** Or a named time zone from the list, for example Europe/Rome. |
| Confirmation tick box | Appears only for hours that are not entirely inside 08:00 to 21:00, including overnight hours. | Required to save those hours. |
Under the hours: "A campaign can use its own hours instead. Hours outside 8:00 to 21:00 are your choice and your
responsibility, and we keep a record of your confirmation."
Under the days: "Customers are called only on these days, in their own local time. Hours that run past midnight count
as the day they start. A campaign can use its own days instead."
### Overnight hours
If the start is later than the end, the hours run overnight, and the card says so, for example: "These hours run
overnight: from 22:00 until 02:00 the next morning." Overnight hours count as the day they start.
### Hours outside 8:00 to 21:00
You can choose hours outside 08:00 to 21:00, for example to call from 07:00. You then have to tick the confirmation,
which reads, for example: "I choose to call customers from 07:00 to 21:00 in their local time. I understand these hours
are outside 8:00 to 21:00 and I am responsible for them."
When you save, RuleInside first checks that you are an owner, then records your confirmation (who, the hours, and
when), and only then saves the hours. Possible messages: "Only an owner of this workspace can confirm calling hours
outside 8:00-21:00." and "Calling hours outside 8:00-21:00 need the owner to confirm that they choose these hours and
are responsible for them."
### How each customer's local time is worked out
RuleInside works out each customer's time zone in this order:
1. the time zone stored on the contact, if there is one;
2. the time zone given by the phone number's prefix, when that prefix points to exactly one time zone;
3. your **Timezone for customers we cannot place**.
If none of these applies, the customer is **skipped**, not guessed. When a number's area spans several time zones, the
call is placed only when your hours hold in all of them.
When you change **Timezone for customers we cannot place** from None to a time zone, the card warns you: "Customers
who are skipped today because we cannot tell their timezone will become callable, in this timezone, once you save."
### Saving
**Save calling hours and days** saves the card. It is greyed out while the hours are not valid or a required
confirmation is not ticked. You see "Calling hours and days saved.", or "Could not save the calling hours. Please try
again." Only owners see this button.
If the card cannot be read, it shows "Could not load your calling hours" and "Please refresh the page. If this keeps
happening, contact support."
If your saved hours or days were ever unreadable, RuleInside calls nobody rather than guessing.
### Campaigns with their own hours
A campaign can set its own hours, its own days, or both. They replace the account's for that campaign only, and hours
and days are replaced independently. They can be changed only while the campaign is still a draft. See
[Campaigns](/campaigns).
## Common questions
Check this page first. If it says **Not confirmed**, an owner must confirm. Then check your calling hours and days.
Only an owner can confirm, withdraw, or change calling hours and days.
No. When only the wording has changed, your calls keep running. Read the new statement and confirm again when you
can.
Consent rules differ by country, so the confirmation must be given again for the new country.
No new calls or messages are started. As the withdrawal statement says, anything already scheduled may still be
delivered.
Yes. Tick **Set my own calling hours**, choose the hours, and tick the confirmation. That choice, and the
responsibility for it, are yours, and RuleInside keeps a record of it.
The customer's. Each customer is called in their own local time.
Their time zone cannot be worked out (set a **Timezone for customers we cannot place**), or it is outside your
hours or days where they are, or they opted out, are on your do-not-call list, have no consent to call, or have
been contacted too often recently. See [How calling rules protect your customers](/outbound-rules).
Partly. WhatsApp **Marketing** templates and a campaign's WhatsApp message after a call wait for the person's
calling hours. Other WhatsApp messages do not. A Marketing follow-up in a chat the customer started goes at once
only within 24 hours of that chat.
No. RuleInside records what you confirm and applies do-not-call, opt-out and calling-time suppression for you.
Whether your lawful basis is valid is for you to judge.
# The campaign page and its results
Source: https://docs.ruleinside.com/campaign-results
Test, launch, pause, resume and cancel a campaign, and read, filter and download every call it placed.
Open a campaign from **Campaigns** (click its name or **View**) to see everything about it: its settings, the
buttons that test, launch, pause, resume and cancel it, the batches of calls it placed and, for a phone campaign,
every call with its result. The page always shows the campaign's current status when you open it.
* **Any member** can read the page, filter the calls, choose columns and use **Download Excel**.
* **Only the account owner** can use **Edit**, **Send test call**, **Launch campaign**, **Pause**, **Resume**,
**Pause again** and **Cancel**. Other members see these greyed out with **Only account owners can manage this: ask
an owner of your team.**
To create a campaign, see [Campaigns](/campaigns).
## The summary
At the top: the campaign name, its status and **← All campaigns**. The summary box shows what the campaign will use.
Fields marked (phone) are not shown for a WhatsApp campaign.
| Field | What it shows |
| - | - |
| **Channel** | **Phone call** or **WhatsApp message**. |
| **Agent** | The agent's name, or "Default agent": your default agent at the time of each call. |
| **Calls from** (phone) | The number, or "Assigned to the agent": the number assigned to that agent is used. |
| **Audience** | **All contacts**, **Dormant > \{days} days**, or **\{count} selected contacts**. |
| **Schedule** (phone) | The start time you set, or **On launch**. |
| **Concurrent calls** (phone) | The saved number, or **Auto** (your plan's maximum). |
| **Ring timeout** (phone) | For example "30s", or **Default**. |
| **Retries** (phone) | For example "up to 3, every 3h (max 2/day)". These are the values actually saved, after any lowering. |
| **Waiting for their calling time** (phone) | Shown only when people are waiting: how many, and **First:** with the earliest time. Each is checked again about an hour before the call. |
| **Calling hours and days** (phone) | **The account calling hours and days**, or the campaign's own, for example "09:00 to 18:00, Mon, Tue, Wed". |
| **Default country code** | The code, or a dash. |
| **Test numbers** (phone) | Your test numbers, or a dash. |
| **Created** | When the campaign was created. |
## The buttons
Which buttons you see depends on the campaign's status. While one action is running, all the buttons are disabled.
| Button | Shown when | What it does | Confirmation |
| - | - | - | - |
| **Edit** | Draft | Opens **Edit campaign**. See [Editing a draft](/campaigns#editing-a-draft). | None |
| **Send test call** | Draft phone campaign with at least one test number | Calls only your test numbers. The campaign stays a draft. | **Place a test call?** then **Place test call** |
| **Launch campaign** | Draft | Starts the campaign for real. | **Launch this campaign?** (phone) or **Send this WhatsApp campaign?** then **Launch now** |
| **Pause** | Running or Scheduled phone campaign | No new call starts. Everyone not yet called keeps their place; retries wait. | **Pause this campaign?** then **Pause campaign** |
| **Resume** | Paused phone campaign | Calling continues inside each person's calling hours and days. | **Resume this campaign?** then **Resume campaign** |
| **Pause again** | Paused phone campaign | Checks every batch again and stops any that is still calling. Safe to press more than once. | None: it runs at once |
| **Cancel** | Draft, Scheduled, Running or Paused, and whenever a batch may still be calling | Ends the campaign for good. | **Cancel this campaign?** then **Cancel campaign** |
In each confirmation, the other button (**Cancel**, **Keep running**, **Stay paused** or **Keep campaign**) closes it
without doing anything.
**What cannot be undone**
* **Launch:** calls already placed and WhatsApp messages already sent cannot be recalled. A launched campaign can
never be launched again or edited.
* **Cancel:** the status becomes **Cancelled** for good. There is no way to restart it. Scheduled batches,
retries and booked call-backs are cancelled with it. Calls already under way cannot be recalled.
* **Pause again** has no confirmation step; it only stops calling, so it is safe to press.
### Send test call
The confirmation reads: "This calls your configured test number(s): \{numbers}: so you can hear the agent before
launching live. It places a real call (standard charges apply) but does not touch your contact list."
A test call is checked against your store's calling consent confirmation, your do-not-call list and opt-outs, this
campaign's calling hours and days, and your daily call limit. It is **not** checked for consent, because a test
number is not one of your contacts: only use your own phones. A test call counts towards your daily call limit.
You then see **Test call placed to \{count} number(s).** or **No test call was placed.** with the reason, for example
that a number is not in international format, is on the do-not-call or opt-out list, is outside the calling hours,
that the daily limit is reached, or that your store has not confirmed its calling consent.
Test calls are for phone campaigns only. To test a WhatsApp template, send it to one number from the
[WhatsApp](/whatsapp) send page.
### Launch campaign
For a phone campaign the confirmation reads: "This places real phone calls to your eligible contacts and will incur
calling charges. We automatically skip anyone who is opted out, on a do-not-call list, without a lawful basis, or
outside their local calling hours. This can't be undone once calls start (you can cancel to stop remaining calls)."
For a WhatsApp campaign: "This sends real WhatsApp messages (your approved re-engagement template) to your eligible
contacts and will incur messaging charges. We automatically skip anyone who is opted out of WhatsApp, on a
do-not-contact list, or without WhatsApp consent. Messages start sending immediately and can't be recalled once
sent."
Each person is called in their own time zone, so the contacts are grouped by time zone, and each group starts at its
next allowed time (never before the schedule you set). A group whose time is more than an hour away waits, and is
checked again, opt-outs included, about an hour before its calls.
A launch can be refused, for example when your account is not active, when the agent has no outbound number, when
the chosen number is no longer available, or, for WhatsApp, when no re-engagement template is enabled. The message
tells you which.
### Pause and Resume
Pause and Resume are for phone campaigns only.
* **Pause:** "No new calls will start. Calls already under way cannot be recalled. Everyone not yet called keeps
their place, and retries wait until you resume." The page then shows **This campaign is paused. No new calls start.
Everyone not yet called keeps their place, and Resume carries on inside each person's calling hours and days.**
* If you see **Campaign paused, but \{count} batch(es) could not be confirmed stopped and may still be calling. Press
Pause again.**, press **Pause again**. If the message repeats, contact support.
* **Resume:** "Calling continues inside each person's calling hours and days. Everyone is checked again for
opt-outs, do-not-call and consent before they are called. Retries that came due while paused go out within about
an hour." Pressing **Resume** does not place a call by itself.
Your account must be active to launch or resume. Pausing and cancelling work even when it is not, so you can always
stop calls.
### Cancel
The confirmation reads: "Any calls already placed can't be recalled, but we'll stop scheduled and in-progress batches
and any retries, so no further contacts are dialed."
If the message says some batches **could not be confirmed** stopped, press **Cancel** again: the button stays
available while any batch may still be calling. If the message repeats, contact support.
## The launch result
Right after a launch or a test call, a box under the summary tells you what happened. It is shown only until you
reload the page.
* **Campaign launched** with, for example, **Dialing \{count} contact(s).** or **\{held} contact(s) are waiting for
their calling time: each is checked again, including opt-outs, about an hour before the call.**
* **Launch complete: nobody was dialed** (phone) or **Send complete: nobody was messaged** (WhatsApp), with **We
matched \{count} contact(s), but none could be called right now.**
* One line per reason people were left out, for example **\{count} skipped: opted out**, **\{count} skipped: phone not
in international format** or **\{count} will be tried again by this campaign when they can next be called**.
| Reason (phone) | Reason (WhatsApp) | What to do |
| - | - | - |
| **opted out** | **opted out** | Nothing: they asked not to be contacted. |
| **on the do-not-call list** | **on the do-not-contact list** | Nothing, while they stay on the list. |
| **no recorded lawful basis to call** | **no recorded consent to message on WhatsApp** | Record their consent in [Contacts](/contacts), or upload the list with a basis. |
| **outside local calling hours** | **outside messaging hours** | Try later, when their calling hours are open. |
| **no timezone on file** | **no timezone on file** | Store a time zone on the contact, or set a default time zone under **Settings › Calling consent**. |
| **called too recently** | **messaged too recently** | Wait: it clears within 7 days. |
| **daily call cap reached** | **daily message cap reached** | Wait: it clears within 24 hours. |
| **a processing error** | **a processing error** | Try again shortly. |
| **your store has not confirmed its calling consent (Settings › Calling consent)** | **your store has not confirmed its messaging consent (Settings › Calling consent)** | The owner confirms it under **Settings › Calling consent**. |
| Not used for calls | **they told WhatsApp to stop marketing messages from your store** | Nothing: only they can turn your marketing messages back on, in WhatsApp. |
People left out at launch because the daily call cap was reached, or because of a processing error, are tried
again by the same campaign, up to one day of your daily call cap per launch, starting an hour later or when their
calling hours next open, for up to about a day, with every check run again. The launch result shows them as
**will be tried again by this campaign when they can next be called**, and anyone beyond one day of your cap, already
being called by another campaign, or whose calling hours are too short for a retry, as **not called by this campaign**. People left out for any other reason, including **called too
recently**, are not called later by that campaign: to reach them, create a new campaign once the reason has passed.
**If nobody was called and nobody is waiting**, the campaign stays a **Draft**: fix the reason and launch it again.
Once at least one person was called or is waiting, it becomes **Running** or **Scheduled** and can never be launched
again. A WhatsApp launch where some messages could not be queued also stays a draft; launching it again reaches the
rest, and nobody receives the message twice. A WhatsApp draft that has already queued messages for some contacts can
be launched again but not edited, even if those messages later failed or were suppressed. Trying to edit it shows
"This WhatsApp campaign has already queued messages for some contacts. It can be launched again to reach the rest,
but not edited."
## Campaign statuses
| Status | What it means | What you can do |
| - | - | - |
| **Draft** | Saved; nobody called yet, or a launch found nobody to call. | Edit, Send test call (phone), Launch, Cancel |
| **Scheduled** | Launched; at least one group is set for a later time. | Pause (phone), Cancel |
| **Running** | Launched and calling (or messaging). | Pause (phone), Cancel |
| **Paused** | Phone only. No new call starts; everyone waiting keeps their place. | Resume, Pause again, Cancel |
| **Completed** | Nothing is left to do. Set automatically. | Nothing |
| **Cancelled** | Stopped for good. | Nothing |
* A phone campaign stays **Running** while it still owes retries or booked call-backs, even after every first call
is done. It also stays open for up to 20 minutes after a batch ends, while a call may still be under way.
* **A WhatsApp campaign keeps showing Running after launch.** Its messages are all queued at launch, but the status
does not change to **Completed** on its own. To see what happened to each message, use the send log on
[WhatsApp](/whatsapp).
## Call batches
A phone campaign hands its calls to the voice provider in batches, one or more per time zone. **Call batches** lists
them, newest first. Before launch: **No batches yet. Batches appear here once the campaign is launched.**
| Column | What it shows |
| - | - |
| **Created** | When the batch was created. |
| **Type** | **First call** (from the launch) or **Call-back** (a later batch placed for this campaign). |
| **Timezone** | The time zone or zones of the batch. |
| **Status** | **Pending**, **In progress**, **Completed**, **Failed**, **Cancelled** or **Cancel failed**. |
| **Scheduled for** | When the batch was set to start, or **Immediate**. |
| **Calls** | How many calls were recorded for the batch. |
| **Results** | Each result with its count, for example "Spoke with the customer 1 · No answer 3". Batches from before 24 September 2026 read **Results are recorded from 24 September 2026**. |
**Cancel failed** (in red) means we asked to stop the batch and could not confirm it stopped: it may still be
calling. Press **Cancel** again.
For a WhatsApp campaign, the table shows **Created**, **Timezone**, **Status**, **Scheduled for** and **Dialed /
Done**, which reads **\{count} message(s) enqueued**.
## WhatsApp after the call
Shown when the campaign has at least one WhatsApp follow-up switched on. **After we spoke** and **We couldn't reach
you** each read **Off**, or:
* **\{count} sent**
* **\{count} waiting** (queued, or held until the calling hours open)
* **\{count} not sent: no WhatsApp permission**
* **\{count} not sent: other reason** (for example an opt-out, the do-not-contact list, the weekly limit, or a
stopped campaign)
* **\{count} failed**
## Calls
For a phone campaign, **Calls** lists every call it placed, one row per call, newest first, 50 per page, with the
number of calls recorded next to the heading.
### Searching and result chips
* **Search a name or phone number**, then **Search**: finds a contact name containing your text (any letter case),
or a number containing the same digits (at least 3).
* Under the toolbar, each result that happened appears as a chip with its count, for example **Spoke with the
customer: 2**. Click a chip to show only those calls; click it again to show them all. You can turn on several
chips; a call matching any of them is shown.
* Extra chips when they apply: **Call-backs requested: \{n}**, **Called us back: \{n}** and **Appointment booked:
\{n}** (only when your agent collects a yes/no field named `appointment_booked`).
* While a filter is on, a line shows each filter (click its × to remove it), **\{count} of \{total} calls match**, and
**Clear filters**.
The page number and the filter are kept in the address bar, so you can reload, bookmark or share a filtered view with
another member of your account.
### Columns of the calls table
| Column | What it shows |
| - | - |
| **When** | Date and time of the call, in your time zone. |
| **Contact** | The contact's name, or a dash. |
| (contact columns) | Any columns you picked in **Columns**. Hover a long value to read all of it. |
| **Phone** | The number called. |
| **Attempt** | 1 for the first call, 2 and 3 for retries. **Requested call-back** (blue) marks the call-back the customer asked for. |
| **Result** | What happened (see below). **Called us back from this number** (green) with a time means a call came in from that number after this call. |
| **Length** | Minutes and seconds, for example 0:41, or a dash. |
| (call data columns) | Any items you picked in **Call data**. Hover a value to read the agent's reason. |
| **Call log** | **Open call** opens the call's page with its transcript and recording. A dash when there is nothing to hear. |
| **Call-back requested by customer** | A dash, or the call-back's state and **Agreed time (your time zone):** with the date and time. |
**Not shown here: calls placed before this list started on 24 September 2026, calls older than 12 months, and calls
to people whose personal data has been erased. Rarely, a call we could not record is also missing.**
Use **Previous** and **Next** to move between pages.
### Call results
| Result | Meaning |
| - | - |
| Spoke with the customer | The customer was reached: they spoke, or the call lasted 15 seconds or more with the agent speaking. |
| Handed to a person | The call was handed over to a person. |
| Customer hung up straight away | Answered and ended at once. |
| No answer | It rang and nobody answered. |
| Line busy | The line was busy. |
| Voicemail | An answering machine picked up. It counts as not reached, and is retried. |
| Wrong number | The number is wrong or invalid. |
| We were not allowed to call | The call was blocked. |
| Problem on our side | A fault on our side or the voice provider's. |
| Cancelled before we called | Called off before the phone rang, for example by a pause or a cancel. |
| Unknown | The voice provider gave no usable result. RuleInside shows Unknown rather than guess. |
| Waiting for the result | The result has not arrived yet. |
### Call-back states
A customer can ask for a call-back during a call when your agent has the call-back booking tool, which our team sets
up. The **Call-back requested by customer** column shows:
| State | Meaning |
| - | - |
| **Booked, not called yet** | Agreed, waiting for its time. |
| **Placing the call** | Being dialled now. |
| **Call placed, waiting for the result** | Dialled; result not in yet. |
| **Called back, answered** | Done. |
| **Called back, no answer**, **line busy**, **went to voicemail**, **the customer hung up straight away** or **wrong number** | Dialled, not reached. An unanswered call-back is not tried again. |
| **Call placed, result unknown** | Dialled; no usable result. |
| **Problem on our side: the person may have been called** | We could not confirm whether the call went out. |
| **Not called** (with a reason) | It was not placed, for example **the person called us first** or **outside the calling hours**. |
| **Cancelled** (with a reason) | It was cancelled, for example **the campaign was cancelled**. |
A booked call-back passes the same checks as every other call when its time comes. If a check fails, it is not
placed and is not tried again.
## Add filter
**Add filter** opens **Filter the calls**. Choose **Filter on**, then **Field**, **Condition** and **Value** as
needed, and press **Apply** (grey until the filter is complete).
| Filter on | Field | Condition and value |
| - | - | - |
| **What the agent collected** | A detail your agent collects, with how many calls have it. | Yes/no: **is yes**, **is no**, **has no value**. Number: **has a value**, **has no value**. Text: **is**, **contains**, **has a value**, **has no value**. |
| **The agent's checks** | A check your agent runs on each call. | **passed**, **failed**, **unknown** |
| **What the agent did** | A tool your agent used. | **used**, **worked**, **failed**, **had an error**, **did not run**, **was blocked** |
| **How the call ended** | (none) | One of the end reasons found on this campaign's calls |
| **Contact column** | One of your contact columns. | **is** or **contains**, then a value |
| **Attempt** | (none) | **First call**, **Retry** or **Requested call-back** |
| **Follow-ups** | (none) | **Called us back** or **Asked for a call-back** |
* Only the kinds this campaign has data for are offered; **Attempt** and **Follow-ups** are always there.
* Every filter must match at once. **is** ignores letter case; **contains** finds the text anywhere.
* Only one **Attempt** and one **Follow-ups** filter at a time: a new one replaces the old one.
* Up to 10 conditions of each kind, values up to 200 characters.
* If you see **This filter could not be used. Clear it to see the calls.**, press **Clear filters**.
## Call data
**Call data** opens a checklist: **Show what the agent collected, its checks and how each call went. Remembered in
this browser. Point at a value to see the agent's reason.** Each ticked item becomes a column.
| Item | What the column shows |
| - | - |
| **Summary** | The call's written summary. |
| **Title** | The call's short title. |
| **How it ended** | The end reason reported by the voice provider. |
| **Language** | The main language, for example "Italian". |
| **Mood** | **Positive**, **Neutral** or **Negative**. |
| **Success score** | The call's success score. |
| **Tools used** | Each tool and its result: **worked**, **failed**, **error**, **did not run**, **blocked**. |
| Each detail your agent collects | **Yes** / **No**, a number or text. Hover for the agent's reason. |
| Each check your agent runs | **Passed**, **Failed** or **Unknown**, with the score when there is one, for example "Passed (8/10)". |
The details and checks offered depend on how your agent is set up, and use your agent's own names. A call without
that piece of information shows a dash.
Your choice is saved in this browser only. A colleague, or you on another browser, starts with nothing ticked.
## Columns
**Columns** (shown when your store has contact columns) opens **Contact columns**: **Show the contact's own details
beside each call. Remembered in this browser.** Each column you tick appears between **Contact** and **Phone**. Like
Call data, the choice is saved in this browser only.
## Download Excel
**Download Excel** (**Preparing the file...** while working) downloads the calls as an Excel file. Any member can
use it.
* It holds what the table currently shows: the active filter, or the whole campaign with no filter. All matching
calls are included, whatever page you are on, newest first, up to **5,000**. If more matched, you see **The file
holds the newest \{count} of \{total} matching calls. Narrow the filter to download the rest.**
* The file is named after the campaign and the date, with one sheet named **Calls**. Times use the time zone of the
browser that downloaded it.
* Columns, in order: **When**, **Contact**, each of your contact columns, **Phone**, **Attempt**, **Was a requested
call-back**, **Result**, **Length**, **Called us back at**, **Call-back requested by customer**, **Call-back agreed
time**, **Summary**, **Title**, **How it ended**, **Language**, **Mood**, **Success score**, then each collected
detail (with its reason), each check (with its score and reason), and **Tools used**.
* If the download fails, you see **The download failed. Please try again. (Nothing was changed.)** and no file is
saved.
## Retries
When a call is not answered, the campaign calls the person again on its own, within the retry settings saved on the
campaign (see **Retries** in the summary).
* A call that connected ends that person's calls: no retry. The exception is a call answered by voicemail, which is
retried.
* Any other ended call (no answer, busy, failed, voicemail) is retried until **Max attempts** is reached.
* The next try is at least **Hours between retries** later, inside the campaign's calling hours and days in the
person's own time zone, and never more than **Max per day** calls on one of their days.
* Every retry passes all the checks again when it is placed. If one fails for a temporary reason (for example the
daily limit), the retry is tried again about an hour later, for up to about a day.
* A paused campaign keeps its retries waiting until **Resume**. A cancelled campaign never retries.
## Common questions
Read the reasons in the launch result. If it says your store has not confirmed its calling consent, the owner
confirms it under **Settings › Calling consent**. **no recorded lawful basis to call** means consent must be
recorded in [Contacts](/contacts). **outside local calling hours** means the people's calling hours are closed
now. **a processing error** means try again shortly. Because nobody was called, the campaign is still a draft and
you can launch it again.
The people are probably **Waiting for their calling time**: the summary shows how many and the first time. Their
calls go out about an hour before that time, after every check runs again.
Only if the first launch called nobody and left nobody waiting (it is then still a draft). Otherwise create a new
campaign. People left out at launch because of the daily limit or a processing error are already tried again by the
first campaign, for up to about a day.
No. Only a draft can be edited. After launch you can pause, resume or cancel it.
No. It is a real call, billed like any other, and it counts towards your daily call limit. It skips the consent
check, so only put your own phones in **Test numbers**.
Calls already under way cannot be recalled. If the message said some batches could not be confirmed stopped,
press **Pause again**.
We could not confirm that the batch stopped. Press **Cancel** again; the button stays available while any batch
may still be calling. If it repeats, contact support.
Look at **Attempt**. A retry is always at least **Hours between retries** after the previous call (minimum 1
hour). A **Requested call-back** is a call the customer asked for, at the time they agreed.
A call came in from that number after our call. Caller ID can be faked, so it means "from this number", not a
confirmed person.
Click **Open call** in the **Call log** column. There is no link when the call has nothing to hear, for example
when nobody answered.
The columns you pick are saved in your browser only. Each person chooses their own.
The file holds up to 5,000 calls, and only those the current filter shows. Calls before 24 September 2026, older
than 12 months, or to people whose data was erased are not in the list or the file.
That is expected: a WhatsApp campaign does not change to Completed on its own. Its messages were queued at launch.
Check each message in the send log on [WhatsApp](/whatsapp).
Look at **WhatsApp after the call**: each follow-up is counted as sent, waiting, not sent (no WhatsApp
permission, or another reason) or failed.
# Campaigns
Source: https://docs.ruleinside.com/campaigns
Call, or send a WhatsApp message to, a group of your contacts: the campaign list and every setting of a new campaign.
A campaign reaches a group of your contacts with the agent and the number you choose. It can place **phone calls**
(your agent speaks to each person) or send a **WhatsApp message** (your approved re-engagement template).
Creating a campaign never contacts anyone: it saves a **draft**. Nothing is dialled or sent until you open the
campaign and press **Launch campaign**. Launching, test calls, pausing, cancelling and reading the results are on
the campaign's own page: see [The campaign page and its results](/campaign-results).
## Who is never contacted
Every person is checked just before each call, retry or message, and anyone who fails a check is skipped:
* people who **opted out**;
* numbers on your **do-not-call** list (on WhatsApp, the do-not-contact list);
* people with **no recorded consent** (a lawful basis to call, or WhatsApp opt-in for a WhatsApp campaign);
* people **outside their calling hours or days**, in their own time zone (phone calls);
* people contacted **too often** recently (at most 5 calls in any 7 days, or 4 WhatsApp marketing messages);
* everyone, when your account's **daily call limit** is reached, or when your store has not confirmed its calling
consent under **Settings › Calling consent**.
If a check cannot be completed, the person is not contacted. The full rules are in
[How calling rules protect your customers](/outbound-rules).
## Who can do what
* **Any member** of your team can open **Campaigns**, see the list, open a campaign, read its results and download
them.
* **Only the account owner** can create, edit, test, launch, pause, resume or cancel a campaign. Other members see
**Only account owners can manage this: ask an owner of your team.** in place of the buttons. See
[Who can do what](/roles).
## The campaign list
The page lists your campaigns, newest first. If your workspace has no store linked yet, it says so and the
**New campaign** button is greyed out.
| Column | What it shows |
| - | - |
| **Name** | The campaign name. Click it to open the campaign. |
| **Type** | **Reactivation**, **Welcome** or **Lead Generation**. |
| **Channel** | **Phone call** or **WhatsApp message**. |
| **Status** | **Draft**, **Scheduled**, **Running**, **Paused**, **Completed** or **Cancelled**. See [Campaign statuses](/campaign-results#campaign-statuses). |
| **Audience** | **All contacts**, **Dormant > \{days} days** (for example "Dormant > 60 days"), or **\{count} selected contacts** for an uploaded list. |
| **Scheduled** | The start time you set, in your own time zone, or **On launch** when you set none. |
| **Created** | When the campaign was created, in your own time zone. |
| (no heading) | **View** opens the campaign, the same as clicking its name. |
With no campaigns yet, the list reads **No campaigns yet. Create one to start reaching your contacts.** If it cannot
be loaded, refresh the page; nothing was changed.
## Creating a campaign
Click **New campaign** (owner only). The form opens with the subtitle **This creates a draft. Nothing is dialed until
you open the campaign and launch it.**
The fields below appear from top to bottom. Fields marked **phone only** disappear when you choose
**WhatsApp message**; anything you typed in them is not saved for a WhatsApp campaign.
### Name
* **Name** (placeholder **Spring win-back**): your own name for the campaign, shown in the list and on its page,
and used in the name of the Excel file of its results.
* Required, 1 to 200 characters. **Create draft** stays grey while it is empty.
### Channel
* **Phone call** (default): **Your agent places a voice call to each contact.**
* **WhatsApp message**: **Sends your approved WhatsApp re-engagement template to each contact who has opted in to
WhatsApp. Configure the template under WhatsApp Templates.** Set the template up first on the
[WhatsApp](/whatsapp) templates page: the launch is refused until one is enabled.
A WhatsApp campaign sends one template message per contact. It has no call-from number, schedule, concurrent calls,
ring timeout, test numbers, calling hours, WhatsApp follow-ups or retry settings, and you cannot upload a list for it
(an uploaded list records consent to be called, which is a different permission from WhatsApp opt-in).
The channel cannot be changed once the draft is saved. To switch between phone and WhatsApp, create a new
campaign.
### Type
* **Type** (placeholder **Choose a type…**): **Reactivation** (default), **Welcome** or **Lead Generation**.
* **A label for your own tracking: it doesn't change who is called (the audience below does).**
### Agent
* **Agent** (placeholder **Choose an agent…**): **The agent that speaks on the call.** Your agents are listed by
name, and your default agent carries **(default)**.
* Default: your default agent. Required. On a WhatsApp campaign the agent is recorded but nobody is called.
### Phone number (call from) (phone only)
* **Phone number (call from)** (placeholder **Choose a number…**): **The outbound number your contacts see.**
* Only your numbers that can place outbound calls are listed. Default: the first of them. Required.
* With no outbound number you see **You have no outbound phone number yet. Add one on the Phone numbers page, then
create a campaign.** Add one on [Phone numbers](/phone-numbers) first.
* The number is checked again when you launch. If it is no longer available for outbound calls, the launch is
refused with **The selected phone number is no longer available for outbound calling.**
### Recipients
* **Recipients** (placeholder **Choose recipients…**) decides who the campaign reaches:
* **Dormant customers** (default): contacts whose last interaction with your agent is older than the number of
days you set below. This is the last interaction, not the last order.
* **All contacts**: every contact in your store that has a phone number.
* **Upload a list (CSV or Excel)** (phone only): exactly the people in a file you upload.
* Whatever you choose, only contacts with a recorded lawful basis, not opted out and inside their calling hours are
called.
**A campaign reaches at most 5,000 contacts.** If the group you choose is larger, the first 5,000 are used and the
rest are not called by this campaign. The **Recipients** field says so. An uploaded list holds at most 1,000 rows:
to reach more people with one campaign, import them on the Contacts page first (up to 1,000 per import, so a
larger list takes several imports; see [Importing contacts](/importing-contacts)), then choose **All contacts**.
Your plan's daily call limit still applies, see [Plans and limits](/limits#daily-calls-for-your-account).
### Dormant for at least (days)
Shown only with **Dormant customers**.
* **Dormant for at least (days)**: **Contacts whose last interaction is older than this.**
* A whole number, 1 or more. Default: 60. Otherwise you see **Enter a number of days (1 or more).** and cannot save.
### Uploading a list
Shown only with **Upload a list (CSV or Excel)**.
* **Recipients (CSV or Excel)**: pick a `.csv` or `.xlsx` file of at most 8 MB. Only the first sheet of an Excel
file is read. An old `.xls` file cannot be read: save it as `.xlsx` or export a CSV.
* **Download CSV template**: a ready-made file with the columns `phone,name,email,city,order_number,language` and
two sample rows.
* **Or paste CSV**: paste the rows instead of uploading a file, for example:
```
phone,name,email
+15551234567,Jane Doe,name@example.com
```
* **A phone column is required (international format, e.g. +15551234567); name, email, and any extra columns become
call variables. This campaign will call exactly these contacts.**
* If the first row names a phone column (for example `phone`, `phone_number`, `mobile` or `Phone Number`), it is
read as a header row. Without a header row, the columns are read as phone, name and email, in that order.
* Every other column becomes a detail your agent can use on the call. A column name may use only letters, numbers
and underscores, start with a letter or underscore, and be at most 64 characters. Columns that break this rule are
listed in red and ignored.
* Rows without a phone number are skipped and counted. The line under the box shows how many recipients are ready
and how many rows were skipped.
* **At most 1,000 rows per list.** Over that, the count turns red and you cannot save: split the file.
**Lawful basis to call** (placeholder **Choose a basis…**): **You must have a lawful basis to call an uploaded list:
we record it with each contact before the campaign can dial.** The bases are web or checkout opt-in, explicit
written consent, verbal consent (recorded call), existing customer relationship, and purchased marketing list. The
last two show a fixed statement and the tick box **I confirm the statement above.**, which you must tick. Choosing a
different basis unticks it.
An uploaded list is imported into **Contacts** when you save, with the basis you chose recorded on each person.
Contacts that already exist with the same number are updated. This import is not undone if the campaign is never
launched, is cancelled, or fails to save. You are responsible for having that basis for every person in the file.
See [Contacts](/contacts).
### Schedule (optional) (phone only)
* **Schedule (optional)**: a date and time, in your own time zone. **Leave empty to start as soon as you launch.
Each contact is still only dialed inside their own local calling hours.**
* A time in the past works like "now".
A schedule does not launch the campaign by itself. You still need to press **Launch campaign**; calls then start
no earlier than the time you set.
### Default country code (optional)
* **Default country code (optional)** (placeholder **+1**): **Used to complete national numbers that aren't already
in +… international format.**
* Shown for both channels. Enter a country calling code, with or without the plus (for example `+44` or `44`).
* A number that still cannot be written in international format is skipped and counted. RuleInside never guesses a
country.
### Concurrent calls (optional) (phone only)
* **Concurrent calls (optional)** (placeholder **Auto**): how many of this campaign's calls may ring or be under way
at the same time. **Blank = your plan's maximum.**
* 1 or more; otherwise **Enter 1 or more.** A number above your plan's maximum is lowered to it when you save. Blank
follows your plan, so an upgrade raises it. See [Plans and limits](/limits#campaigns).
### Ring timeout (optional) (phone only)
* **Ring timeout (optional)** (placeholder **seconds**): **How long to ring (5–300s).**
* 5 to 300 seconds; otherwise **5–300 seconds.** Blank uses the voice provider's default.
### Test numbers (optional) (phone only)
* **Test numbers (optional)** (placeholder **+15551234567**): **Your own numbers, one per line. You can place a test
call to these before launching live.**
* One number per line, in international format (or a national number with a default country code). At most 20.
Extensions and pauses written into a number are refused.
* **Send test call** appears on the campaign page only for a phone draft with at least one test number.
Only put your own phones here, never a customer's number. A test call skips the consent check, and it is a real,
billable call.
### Calling hours and days (phone only)
The panel shows the hours this campaign will use, for example **This campaign uses the account calling hours: 08:00
to 21:00, every day, in each customer's local time.** Your account's hours start as 08:00 to 21:00, every day, and
are changed under **Settings › Calling consent**.
* **Use its own calling hours and days instead of the account's** (unticked by default): gives this campaign its own
hours.
* Two hour pickers, start (00:00 to 23:00) **to** end (01:00 to 24:00). They start at your account's hours. The
start and end cannot be the same hour.
* Seven day buttons, Monday first. At least one day must stay selected.
* Hours that are not entirely inside 08:00 to 21:00 need a second tick: **I choose to call from \{start} to \{end} in
each customer's local time. I understand these hours are outside 8:00 to 21:00 and I am responsible for them.**
Who confirmed, when, and which hours are recorded.
* **Each customer is called in their own local time. Retries follow the same hours and days. Hours that run past
midnight count as the day they start.**
### WhatsApp after the call (phone only)
**Optional. Sent once, when each person's calling ends, only to people who agreed to receive WhatsApp messages from
you, and only inside the calling hours.**
* **After we spoke**: sent to a person who was reached.
* **We couldn't reach you (after the last attempt)**: sent to a person who was not reached, after their last
attempt.
Each defaults to **Off**. The other choices are your enabled WhatsApp message templates, shown as name and language
code. Each blank in the template must be filled from the contact's own details; otherwise the template is refused.
A template that was disabled or deleted after you chose it shows as **A template that is no longer available (choose
another, or Off)**.
The weekly WhatsApp limit, opt-outs and the do-not-contact list still apply. A follow-up from a paused campaign
waits for **Resume**, and one that still cannot be sent after 7 days is dropped.
### Show retry settings (phone only)
Click **Show retry settings** to open the panel (and **Hide retry settings** to close it). **If a call isn't answered
we retry, within these limits. We cap retries to protect your contacts (and you) from over-calling, so larger values
are reduced automatically.**
| Field | Default | What it does | Allowed |
| - | - | - | - |
| **Max attempts (up to 3)** | 3 | Total calls to one person, the first call included. | 1 to 3 |
| **Hours between retries (min 1)** | 3 | The shortest wait after an unanswered call. | 1 to 168 (one week) |
| **Max per day (up to 2)** | 2 | The most calls to one person on one of their local days. | 1 to 2 |
A blank or 0 uses the default. A value outside the range is changed to the nearest allowed value when you save,
without a message: the campaign page shows the values actually saved, under **Retries**. How retries work is
explained in [Retries](/campaign-results#retries).
### Saving
* **Cancel** closes the form. Nothing is saved.
* **Create draft** (**Creating…** while working) saves the draft. It stays grey until every required field is valid.
On success you see **Campaign created as a draft: review it, then launch.** and the campaign opens.
If you chose your own calling hours and only that part fails to save, the draft is still created and uses your
account's hours; the message says so.
## Editing a draft
On the campaign page, **Edit** (owner, drafts only) opens **Edit campaign**: **You can change a campaign until it is
launched. Nothing is dialed until you launch it.**
* The same fields, filled in with the saved values. **Channel** is locked.
* **Recipients** offers **Keep the current list (\{count} contacts)** for an uploaded list. Choosing **Upload a list
(CSV or Excel)** again imports a new list and replaces the audience.
* Emptying an optional field returns it to its default (**On launch**, **Auto**, the provider's ring time, no test
numbers).
* **Save changes** (**Saving…**) saves. On success: **Campaign saved. Review it, then launch.**
Once a campaign has been launched it cannot be edited: it keeps the settings it was launched with. After launch you
can only pause, resume or cancel it.
There is no way to delete a campaign. A draft you no longer want can be cancelled.
## Common questions
**Create draft** only saves a draft. Open the campaign and press **Launch campaign**. A schedule does not launch
it for you either.
Up to 5,000. A larger group is cut to the first 5,000. An uploaded list holds at most 1,000 rows, so for more
people either run several campaigns, or import them on the Contacts page (1,000 per import) and choose **All
contacts**. Your plan's daily call limit, from 500 calls in 24 hours on Lite to 15,000 on Scale, still applies:
a launch calls up to what is left of it, queues at most one more day of it to try again later, and does not
call anyone beyond that; the launch result says how many. See
[Plans and limits](/limits#daily-calls-for-your-account).
Mostly. People left out at launch because your daily call cap was reached, or because of a processing error,
are tried again by the same campaign, up to one day of your daily call cap per launch, starting an hour later
or when their calling hours next open, for up to about a day. Every check runs again before each call. The launch result says how many will be
tried again and how many will not: people beyond one day of your cap are not called by that campaign, so launch
another campaign later to reach them. See [The campaign page](/campaign-results#the-launch-result).
Retries are capped: at most 3 attempts, at least 1 hour apart, at most 2 a day. Larger values are lowered when
you save.
The number is lowered to your plan's maximum when you save. See [Plans and limits](/limits#campaigns).
**Dormant customers** uses the last time the contact interacted with your agent. The **Win back past customers**
automation uses the date of the last order in your store. They can select different people. See
[Automations](/automations).
Yes, on one campaign: tick **Use its own calling hours and days instead of the account's**, set the hours, and
confirm that you are responsible for hours outside 8:00 to 21:00. Your confirmation is recorded with who made it
and when. Each person is still called in their own local time.
No. The channel is fixed once the draft is saved. Create a new campaign.
Only the account owner can. Other members can view campaigns, filter the calls and download the results.
No. The contacts stay in **Contacts** with the lawful basis you recorded. Manage them from
[Contacts](/contacts).
# Connections
Source: https://docs.ruleinside.com/connections
Connect your store, and let your own systems send contacts, events and do-not-call numbers.
Open **Connections** in the sidebar. The page has two cards:
* **Connect your store**: lets your agent look up orders, products and shipping in your online store.
* **Inbound connections**: lets your other systems (a CRM, a calendar, a booking or form tool, or your own code)
send contacts and events **into** RuleInside.
To send data **out** of RuleInside to your own systems, use [Webhooks](/webhooks) instead.
## Who can do what
* **Everyone on your team** can see whether the store is connected, see the inbound connections, and use **Copy
URL**.
* **Only the owner** can connect the store and create, switch on or off, change the secret of, or delete an inbound
connection. Members see a note that only an account owner can do this.
See [Who can do what](/roles).
## Connect your store
You never copy or paste a key. You approve the connection in your own store's admin, your store hands the access
keys to us directly, and we check them against your store before anything is saved.
The first line of the card says which kind of store your account is set up for. This is set when we set up your
account; there is no store type to choose on this card.
| What the card says | What it means |
| - | - |
| "Your account is set up for a WooCommerce store." | You can connect your store yourself, as below. If your store runs on FluentCart instead, contact us before connecting. |
| "Your account is set up for a FluentCart store." | For now our team connects FluentCart stores. Contact us and we connect yours with you. |
| "Store connection is available for WooCommerce and FluentCart stores." | Contact us to connect a store on another platform. |
### Fields and buttons
| Item | What it does |
| - | - |
| Status line | "Connected to" and your store's checked address, or "No store is connected yet." |
| Last attempt line | The result of your latest connection attempt. See [What the last attempt says](#what-the-last-attempt-says). |
| **Your store address** | Your store's public address, for example `https://shop.example.com`. Up to 2,048 characters. Filled in with the connected address when there is one. |
| **Connect your store** | Starts the connection (the first time). Available once an address is typed. While it works it reads "Opening your store...". |
| **Connect again** | Replaces **Connect your store** once a store is connected. Available once you tick the warning box below. |
| **Check again** | Reloads the card to show the latest status. It does not contact your store. |
### Connect a WooCommerce store
Use the public `https` address of your store.
Your store's own WooCommerce approval page opens in a new tab. Approve the app named RuleInside (followed by
your account id) with **read and write** access. WooCommerce offers no narrower choice, and your agent makes
some changes in your store.
Your store sends the access keys straight to us, and we test them against your store before saving. The card
shows the result.
The approval page is valid for **10 minutes**. You can start at most **5 connections in 10 minutes**.
### Connecting again
**Connect again** creates a new access key in your store, and the old key keeps working until you delete it in
your store's admin. Before you can connect again, tick "I understand, and I will delete the old key in my store's
admin." Then delete the old key in WooCommerce afterwards.
### What the last attempt says
| On the card | What it means |
| - | - |
| "Last connection: approved in your store and checked by us." | Connected and checked. |
| "Waiting for you to approve the connection in your store. The approval page stays valid for 10 minutes." | You started but have not approved yet. |
| "You approved the connection. We are checking it against your store now." | Our check is running. You cannot start again meanwhile; click **Check again** in a minute. |
| "The last approval page expired before it was used. Start again when you are ready." | The 10 minutes passed. |
| "The connection was approved on a different website than the store address entered here. Check the address and start again." | You approved it on another site. |
| "Your store did not give us the access we asked for. Start again and approve the access your store asks you about." | The access was refused, or less than read and write was approved. |
| "We could not reach your store to check the connection. Check that your store is online, then start again." | Your store was offline or did not answer as expected. |
| "Your store did not send the access details. Start again and complete the approval in your store." | The approval was not finished. |
| "Your store is not the kind your account is set up for. Contact us." | The store is a different platform from the one your account is set up for. |
| "This store is already connected to another RuleInside account, so it was not connected here. Contact us if it is yours." | Another account has this store. |
| "Something went wrong on our side while connecting. Start again, or contact us if it happens again." | The problem was on our side. |
| "We could not load your store connection. Refresh the page to try again." | The status could not be read. You cannot connect until it can; refresh the page. |
### Messages when you click Connect
| Message | What to do |
| - | - |
| "Enter your store's address." | Type your store's address. |
| "Confirm the warning about the old key first." | Tick the warning box. |
| "That store address cannot be used. Enter the public https address of your store." | Use your store's public `https` address. |
| "We could not find that store address right now. Check it and try again in a minute." | Check the address for typing mistakes, then try again. |
| "You already approved a connection and we are still checking it. Press Check again in a minute." | Wait, then click **Check again**. |
| "Too many attempts in a short time. Wait ten minutes, then try again." | Wait ten minutes. |
| "Your store is not the kind your account is set up for. Contact us." | Contact us. |
| "For now our team connects this kind of store. Contact us and we will connect it with you." | Contact us. |
| "Store connection is not available for your store's platform." | Contact us. |
| "We could not check your store connection. Please try again." | Try again in a moment. |
| "This store is already connected to another RuleInside account. Contact us if it is yours." | Contact us. |
| "We could not start the connection. Please try again." | Try again; contact us if it keeps happening. |
## Inbound connections
An inbound connection gives another system its own web address (a **webhook URL**) and a **secret** to send you
contacts and events. You then say where each field is in the data that system sends.
Incoming data never places a call on its own and never gives calling consent. A call only happens through an
[Automation](/automations), and always passes the consent and calling-hours checks.
There is no fixed list of systems to choose from: any system that can send a secure (HTTPS) request with JSON data
and a custom header can send you data. You name each connection yourself, and the field mapping adapts it to the
system that sends.
### Create a connection
Click **New connection** (owner only). Fill in:
| Field | Rules | What it does |
| - | - | - |
| **Connection id** | Required. Lowercase letters, numbers and hyphens, starting with a letter or number, up to 40 characters. Must be unique in your account. | Names the connection, for example `hubspot` or `calendly`, and becomes the end of its webhook URL. It cannot be changed later. |
| "The sender already uses our event format (no mapping needed)" | Off by default | Switch it on if the sender already sends events in RuleInside's own format (see [Our event format](#our-event-format)). The mapping fields are then hidden, except **Do not call - path**. |
| "Switch on now" | On by default | Whether the connection accepts data as soon as it is created. |
With the format switch off, tell us where each value is in the sender's data, using **dot paths**. `customer.email`
means "the email inside customer"; `contacts.0.email` means "the email of the first contact". Leave blank what the
sender does not send. Each path can be up to 200 characters.
| Field | Example | What it does |
| - | - | - |
| **Event type - path** | `action` | Where the sender's data says what kind of event it is. If filled in, the fixed type is ignored. |
| **Event type - fixed** | `lead.created` | One event type used for every message, when the sender does not say. Used only if there is no path. |
| **Email - path** | `customer.email` | Where the contact's email address is. |
| **Phone - path** | `customer.phone` | Where the contact's phone number is. |
| **Name - path** | `customer.name` | Where the contact's name is. |
| **Their record id - path** | `customer.id` | The contact's id in the sender's system, used to recognise the same person next time. |
| **Do not call - path** | `customer.do_not_call` | Where the sender marks a contact not to be called. See [Do-not-call from your own system](#do-not-call-from-your-own-system). |
| **Their event id - path** | `id` | The sender's own id for each event, so a message sent twice is recorded once. |
Click **Create connection** (available once a **Connection id** is typed). If the format switch is off and every
mapping field is blank, you are asked to map at least one field or switch on the format option. If the id is
already used, you see that a connection with that name already exists.
On success you see "Connection created." and the credentials panel.
### Copy the URL and the secret
After you create a connection, or after **New secret**, a yellow panel shows:
* **Webhook URL**, with a **Copy** button;
* **Header (X-RI-Ingest-Secret)**: the secret, with a **Copy** button. The sending system must send it in an HTTP
header named `X-RI-Ingest-Secret` with every request.
Put both into the system that will send you data, then click **Done**.
The secret is shown only once. After you click **Done** it cannot be shown again; the only way to get a usable secret
is **New secret**.
### Each connection on the list
The card shows how many connections you have. Each one shows:
* its **Connection id** and a badge: **Active** (switched on) or **Off**;
* its webhook URL;
* an **On** / **Off** switch (owner only). It takes effect straight away, with no save button. If the change
fails, the switch flips back and you see "Could not update the connection.";
* **Copy URL** (everyone): copies the webhook URL;
* **New secret** (owner only);
* **Delete** (owner only).
**New secret** asks you to confirm ("Replace the secret?"), then replaces the secret at once. The old secret stops working immediately,
and the sending system is refused until you update its `X-RI-Ingest-Secret` header with the new secret. This cannot
be undone.
**Delete** asks you to confirm ("Delete this connection?"); choose **Delete** or **Keep it**. The webhook URL stops
accepting data, and this cannot be undone. Contacts and events it already delivered are kept. You can create a
connection with the same **Connection id** again later; it gets a new secret.
If you see "Couldn’t load your connections. Refresh to try again.", the list could not be loaded at that moment,
even if the count says you have none. Refresh the page.
### Do-not-call from your own system
A connection can say where the sender marks a contact **not to be called** (**Do not call - path**). A clear yes
adds the number to your do-not-call list, so your campaigns never call it. These values count as yes, in any mix
of capital and small letters: `true`, `yes`, `y`, `1`, `si`, `sì`, `sí`.
The number blocked is the one in the message, together with every number already stored for a contact matched by
their record id or email.
Nothing sent this way removes a number from your do-not-call list. A connection can only add numbers.
### Our event format
If you switch on "The sender already uses our event format (no mapping needed)", each message must be a JSON object
like this one:
```json Event theme={null}
{
"type": "lead.created",
"event_id": "evt-123",
"contact": {
"email": "name@example.com",
"phone": "+15555550100",
"external_id": "crm-456",
"name": "Sample Customer"
},
"payload": { "source": "contact form" }
}
```
```bash Sending it theme={null}
curl -X POST "https://your-webhook-url-from-the-portal" \
-H "Content-Type: application/json" \
-H "X-RI-Ingest-Secret: your-secret" \
-d @event.json
```
* `type` is required: 1 to 64 letters, digits, `.`, `_` or `-`.
* `event_id`, `occurred_at` and `payload` are optional.
* `contact` needs at least one of `email`, `phone` or `external_id`. It can also carry `name` and `attributes`.
The same rules apply to a mapped message once it has been mapped: it needs an event type and at least one of email,
phone or record id.
### What the sending system gets back
| Answer | Meaning |
| - | - |
| Success | The event was recorded. |
| **401** | Unauthorized: the connection is unknown or switched **Off**, or the secret is wrong or missing. All of these look the same on purpose, so outsiders cannot find out which connections exist. |
| **400** | The data is not valid JSON, or after mapping the event type is missing or not valid, or the contact has no email, phone or record id. |
| **408** | The message arrived incomplete. It is safe to send it again. |
| **413** | The message is larger than 128 KB. |
| **502** | A do-not-call instruction could not be recorded. The sender should try again. |
## Common questions
No. You approve the connection in your store, and your store sends the access to us directly.
That is expected: the old key is not removed automatically. Delete the older one in your store's admin.
You have not approved the connection yet, or you closed the approval tab. The approval page is valid for 10
minutes; start again if it has expired.
For now our team connects FluentCart stores. Contact us and we connect yours with you.
Any system that can send a secure (HTTPS) request with JSON data and a custom header. There is no fixed list.
Check, in order: the connection is **Active**; the header name is exactly `X-RI-Ingest-Secret`; the secret is
the latest one (after **New secret** the old one no longer works); and the whole URL was copied.
Click **New secret** and update the sending system with the new one. The old secret cannot be shown again.
No. Calls only happen through an [Automation](/automations), after the consent and calling-hours checks.
No. A connection can only add numbers to it.
# Contacts
Source: https://docs.ruleinside.com/contacts
Your customer list, their consent on each channel, and who must never be contacted.
**Contacts** is your store's list of customers for outbound calls and WhatsApp messages, with consent and opt-outs
built in. The rule that governs the whole page is printed at the top:
> A contact can be called or messaged only once you record a lawful basis. Do-not-call numbers, opt-outs and
> out-of-hours calls are suppressed automatically.
When you have contacts, the same line ends with the total, for example "· 36 contacts".
Contacts belong to your whole store, not to one agent, so this page has no agent selector.
## Two words used on this page
* **Lawful basis:** the reason you are allowed to contact a person, for example "they ticked a box at checkout" or
"they are an existing customer". You record which basis you rely on for each contact. RuleInside does not check
that the basis is true; you are responsible for it.
* **Consent:** on a contact, consent means you have recorded a lawful basis **for one channel**. Phone calls and
WhatsApp are recorded separately, and one never implies the other.
## How a contact becomes reachable
A newly imported contact cannot be called or messaged. It becomes reachable on a channel only once a lawful basis is
recorded for that channel. Even then, every call and every message is checked at the moment it is sent:
1. your store's [Calling consent](/calling-hours) confirmation;
2. the person has not opted out;
3. the number is not on your do-not-call list;
4. consent is recorded for that channel;
5. it is inside your calling hours (calls only);
6. the person has not been contacted too often, and your account is within its daily limit.
If any check fails, the person is not contacted. See
[How calling rules protect your customers](/outbound-rules).
## Who can do what here
| Action | Member | Owner |
| - | - | - |
| See the list, search it, page through it | Yes | Yes |
| Choose which columns you see, and their order | Yes | Yes |
| **Opt out** a contact | **Yes** | Yes |
| **Import contacts**, **Import with consent** | No | Yes |
| **Edit** a contact | No | Yes |
| Record or update consent (**Consent**) | No | Yes |
| **Add column**, **Manage fields** | No | Yes |
| See and decide [store accounts to review](#store-accounts-to-review) | No | Yes |
Opting someone out is open to every member on purpose: honouring a "stop contacting me" request should never wait for
an owner.
A member sees this where the import buttons would be: "Only account owners can import contacts or change consent -
ask an owner of your team." If a member tries an owner action anyway, it is refused with "Permission denied: this
action requires an account owner (billing) permission."
## The buttons at the top
* **Columns** opens the Columns panel, where you choose what you see and owners manage the workspace's custom fields.
When columns are hidden, the button shows how many, for example "Columns -2". See
[Showing, hiding and adding columns](/importing-contacts#showing-hiding-and-adding-columns).
* **Import contacts** imports a CSV. The people arrive as **Calls: no consent**. Owners only.
* **Import with consent** imports a CSV and records a lawful basis to call for everyone in it. Owners only.
Both import buttons are greyed out until a store is linked to your workspace. Everything about importing is on
[Importing contacts](/importing-contacts).
## The list
The list shows **50 contacts per page**, newest first. When there is more than one page, "Page 1 of 3" appears with
**Previous** and **Next** (greyed out on the first and last page). The header row stays visible while you scroll.
| Column | What it shows | Can you hide it? |
| - | - | - |
| **Name** | The contact's name, or a dash if there is none | Yes |
| **Phone** | The phone number as stored, or a dash | No. The phone number is the contact's key: consent and the do-not-call list are tied to it. |
| **Email** | The email, or a dash | Yes |
| **Consent** | The consent badges, and under a **Calls: callable** or **WhatsApp: opt-in** badge a small line with the basis and the date, for example "explicit written · 27 Sept 2026" | No |
| One column per custom field | The value stored for that contact, or a dash. The header is the field name, for example `lead_interest`. | Yes. Hidden by default when no contact on the page has a value. |
| **Last contacted** | The date of the last outbound call attempt recorded for that contact, or a dash | Yes |
| Actions | The buttons for that contact | No. Always last, and pinned to the right edge so you can reach it when the table scrolls sideways. |
**Contacts without a phone number** can appear in the list. They come from your store's orders and customers, from
conversations, or from other sources. They cannot be called, granted consent or opted out from this page; owners can
only edit them.
### Searching
The **Search this page** box appears when there are contacts. It searches **only the contacts on the current page**
(up to 50), by name, phone, email and calling basis. To find someone on another page, move to that page with
**Previous** or **Next** and search again.
### When the list is not shown
| What you see | What it means |
| - | - |
| "This workspace isn't linked to a store yet." | No store is connected. The import buttons are greyed out. |
| "We couldn't load your contacts right now. Please refresh - if it keeps happening, contact support. (Nothing was changed.)" | The list could not be read. This is not the same as having no contacts. Refresh the page. |
| "No contacts yet. Import a CSV to get started." | Your store has no contacts yet. |
| "This page is empty - use Previous to go back." | The page number is past the end, for example from an old bookmark. |
| A message that no contacts on this page match your search | Your search filtered out every contact on the current page. |
## Store accounts to review
Sometimes a shopper creates or updates an account on your store with an email address that matches an older
contact who is not linked to any store account. RuleInside does not link them on its own: it shows the owner a
suggestion above the list, titled for example **1 store account to review**. Only owners see it, and only when there
is something to decide.
Each suggestion shows the contact (name, email and phone) and a line such as "Store account #123, seen" followed by
the date the match was first seen. You have two buttons:
* **Link** opens **Link this contact to the store account?**, which explains what linking does. Click **Yes, link
them** to link, or **Cancel**. You see "Contact linked to the store account."
* **Not the same person** dismisses the suggestion without changing the contact. You see "Marked as not the same
person." A dismissed suggestion does not come back.
Link only if you are sure it is the same person. Once linked, a shopper logged in to that store account is treated as
this contact: your agent greets them by this contact's name, can tell them this contact's order history, and sees
this contact's past call summaries, notes and preferences. Later changes to the account's name and email also update
this contact.
If linking is refused, nothing is changed and you see why, for example "This store account is already linked to
another contact. Nothing was changed." or "This contact is already linked to a store account. Nothing was changed."
## Consent badges
Each contact shows a **Calls** badge. A **WhatsApp** badge appears only when there is something to say about
WhatsApp.
| Badge | Colour | What it means |
| - | - | - |
| **Calls: no consent** | Grey | No lawful basis to call is recorded. The contact cannot be called. Every contact starts here after a plain import. |
| **Calls: callable** | Green | A lawful basis to call is recorded and the person has not opted out of calls. Each call is still checked against your do-not-call list, calling hours and limits. |
| **Calls: opted out** | Red | The person will never be called again. This wins over everything else: it shows even if a basis to call was recorded at some point. |
| **WhatsApp: opt-in** | Dark green | A lawful basis to message on WhatsApp is recorded and the person has not opted out of WhatsApp. |
| **WhatsApp: opted out** | Red | The person will not receive WhatsApp messages. |
| No WhatsApp badge | | No WhatsApp consent and no WhatsApp opt-out on record. Most contacts look like this. |
The two channels are independent. **Calls: callable** does not allow WhatsApp messages, and **WhatsApp: opt-in** does
not allow calls.
## The buttons on each contact
On a wide screen each contact shows up to three buttons. On a narrow screen, such as a phone, they collapse into one
"..." button that opens a menu with the same choices, with **Opt out** in red at the bottom.
| Button | Who sees it | When it shows | What it does |
| - | - | - | - |
| **Edit** | Owners | Always | Opens **Edit contact** |
| **Consent** (in the narrow menu: **Grant consent**, or **Update consent** when a basis to call is already recorded) | Owners | The contact has a phone, and at least one channel is still open (not opted out) | Opens **Record consent** |
| **Opt out** (red) | Every member | The contact has a phone and is not already opted out of calls | Opens **Opt this contact out?** |
So a contact with no phone shows only **Edit**, and only to owners.
## Recording consent for one contact
Use **Record consent** to record, or replace, the lawful basis for one contact on calls or on WhatsApp. Owners only.
The dialog shows the contact's name and number, with "record the lawful basis on which your agents may contact this
person on the channel you choose. You are responsible for having that basis."
| Field | What it does | Options and rules |
| - | - | - |
| **Channel** | Choose **Phone calls** or **WhatsApp**. | Shown only when both channels are open for this contact. A channel the person opted out of is never offered. The dialog opens on calls when calls are open. Switching channel clears the basis and the tick. |
| **Lawful basis to call**, or **Lawful basis to message on WhatsApp** | Choose the basis. Starts as "Choose a basis…". | The list depends on the channel. See [Lawful bases](#lawful-bases). |
| **I confirm the statement above.** | Appears under a fixed statement for the two bases that need one. | Must be ticked. Changing the basis clears it. |
**Buttons:** **Cancel** closes without saving. **Record consent** saves (it shows "Saving…" while it runs). It stays
greyed out until you choose a basis and tick any required statement.
**When you save,** the basis and today's date are recorded for that channel, replacing any earlier basis on that
channel, and a record of who did it is kept. You see one of these:
* "Consent recorded - this contact can now be called."
* "Consent recorded - you can now message this contact on WhatsApp."
* "No change (the number could not be recorded)." Nothing was written.
* "Could not record consent." or a more specific message, for example when the stored number cannot be put in
international format.
You may also see these messages inside the dialog:
* "There is no channel available for this contact right now - they have opted out, or the lawful-basis list could
not be loaded." Refresh the page; if the person has opted out of both channels, consent cannot be recorded.
* "These are the options for a different channel, so consent cannot be recorded here right now. Please reload the
page, and contact support if it keeps happening." The save button stays greyed out.
There is no button that removes consent while leaving the contact reachable. **Update consent** replaces the basis;
it cannot set a contact back to **Calls: no consent**. To stop contacting someone, use **Opt out**.
### Lawful bases
| Basis, as shown in the picker | Calls | WhatsApp | Statement to confirm | Can you choose it? |
| - | - | - | - | - |
| Web / checkout opt-in | Yes | Yes | No | Yes |
| Explicit written consent | Yes | Yes | No | Yes |
| Verbal consent (recorded call) | Yes | No | No | Yes |
| Existing customer relationship | Yes | No | **Yes** | Yes |
| Purchased marketing list | Yes | No | **Yes** | Yes |
| WhatsApp user-initiated (messaged us first) | No | Yes | No | Yes |
| WhatsApp opt-in given on a recorded call | No | Yes | No | No. RuleInside records it when a person says yes to WhatsApp messages on a campaign call. |
| Replied START on WhatsApp | No | Yes | No | No. RuleInside records it when a person who had opted out of WhatsApp replies START. |
So on WhatsApp you can choose from three bases, and none of them needs a statement. The two bases with a statement
are for calls only, because the statement is about phone calls.
Under each **Calls: callable** or **WhatsApp: opt-in** badge, the small line shows the basis in short form (for
example "purchased list" or "wa user initiated") and the date it was recorded.
"I confirm that these contacts are existing customers, or people with whom I have a prior relationship that gives
me a lawful basis to contact them by phone using an automated AI voice agent, and that I am solely responsible for
the lawfulness of contacting them. I understand that RuleInside applies do-not-call, opt-out, and calling-time
suppression on my behalf."
"I confirm that I have a lawful basis to contact these people by phone using an automated AI voice agent, that the
contacts were obtained in a way that permits such calls, and that I am solely responsible for the lawfulness of
contacting them. I understand that RuleInside applies do-not-call, opt-out, and calling-time suppression on my
behalf, but does not verify how I obtained these contacts or whether I may lawfully call them."
The statement you confirm is stored word for word, together with who confirmed it and when.
## Editing a contact
**Edit** opens **Edit contact**, where an owner corrects a contact's details and fills in their custom fields. The
custom columns are passed to your agent on an outbound call as variables.
| Field | Can you change it? | Notes |
| - | - | - |
| **Name** | Yes | Free text. |
| **Email** | Yes | An email can belong to only one contact in your store. |
| **Phone** | No (greyed out) | "The phone number is the contact's key (consent & do-not-call are tied to it). To change it, opt this contact out and re-import under the new number." |
| One box per custom field, labelled with the field name | Yes | Leave a box blank to clear that value. |
**Buttons:** **Cancel** closes without saving. **Save** saves (it shows "Saving…"). You see "Contact saved.", or
"Could not save the contact." if it failed.
Values for fields that are not current workspace columns are kept as they are. To add a column, see
[Workspace fields](/importing-contacts#workspace-fields-owners-only).
## Opting a contact out
Use **Opt out** when a person asks not to be contacted. Any member can do it.
The confirmation, **Opt this contact out?**, reads: "+15551234567 will be added to your do-not-call list for both
calls and WhatsApp, and any consent on either channel will be revoked. This suppression survives a re-import and
cannot be reversed from the portal. On WhatsApp, the customer can opt back in themselves by replying START."
**Buttons:** **Cancel** closes without changing anything. **Opt out** opts the contact out straight away.
**What happens when you click Opt out:**
* The contact is opted out of **calls and WhatsApp**, and any consent on either channel is removed.
* The number is added to your do-not-call list for calls and for WhatsApp.
* Any calls already scheduled for the contact are cancelled, and any WhatsApp messages waiting to be sent to them are
stopped.
* A record is kept for each channel.
* You see "Contact opted out - they will not be contacted again on calls or WhatsApp."
If the stored number is not in international format, the opt-out cannot be completed and you see: "We couldn't opt
this contact out - the stored phone number isn't in international format. Please edit the contact to a +… number
and try again." The phone number cannot be edited in the portal, so if you see this message, contact support.
Other failures show "Could not opt this contact out."
**An opt-out cannot be undone by you.** Not with a button, not by importing the number again (an import never
touches opt-outs or the do-not-call list), and not by recording consent again (the opt-out still wins, and the badge
stays **Calls: opted out**). Only the customer can reverse it, and only for WhatsApp, by replying START. See
[WhatsApp](/whatsapp).
The **Opt out** button stays until the contact is opted out of **both** calls and WhatsApp. If the person opted out
of calls only, for example by saying so during a call, **Opt out** stops WhatsApp too.
## Your do-not-call list
Your store has a do-not-call list, kept separately for calls and for WhatsApp, and checked before every call and every
message. There is no page in the portal to view or edit it. Numbers are added to it when:
* someone uses **Opt out** on a contact (calls and WhatsApp);
* a person says during a call that they do not want to be called (calls);
* a person replies **STOP** on WhatsApp (calls and WhatsApp). The whole message must be the word STOP, in any letter case,
or a quick-reply button that says STOP. "Stop calling me" does not count;
* your own system sends a "do not call" request through a [Connection](/connections) (calls);
* your system uses the RuleInside API.
A STOP on WhatsApp stops both calls and WhatsApp. An opt-out from calls made during a call does not stop WhatsApp.
## Common questions
A plain import adds people as **Calls: no consent**. Record a lawful basis with **Import with consent**, or with
**Consent** on each contact. If the badges say **Calls: callable** and still nobody is called, check
[Settings, Calling consent](/calling-hours): until an owner confirms it, nothing is sent. Then check your calling
hours and days.
No, only owners. Any member can opt a contact out.
No. Not with a button, not by importing the number again, and not by recording consent again. On WhatsApp, the
customer can opt back in by replying START. For calls there is no way back.
Yes. **Opt out** stops both calls and WhatsApp, and cancels anything already scheduled for that contact on either
channel.
Use **Opt out** on that contact: it is still shown while WhatsApp is open, and it stops WhatsApp too. The customer
can also reply **STOP** on WhatsApp, or your own system can opt them out through the RuleInside API.
No. Each channel needs its own lawful basis.
That basis, and Purchased marketing list, are for calls only, because their statement is about phone calls. On
WhatsApp you can choose Web / checkout opt-in, Explicit written consent, or WhatsApp user-initiated (messaged us
first).
Not in the portal. **Update consent** changes the basis but cannot remove it. To stop contacting someone, use
**Opt out**.
No. You are responsible for it. RuleInside records who recorded which basis and when and, for the two bases with
a statement, the exact statement that was confirmed.
You cannot edit it, because consent and the do-not-call list are tied to the number. If needed, opt the old number
out, then import the new number. Always write numbers in international format with a `+` and no spaces, for
example `+15551234567`.
Search covers only the current page of 50 contacts. Use **Next** to move through the pages.
They came from your store's orders and customers, or from conversations. They cannot be called, given consent or
opted out from this page. Owners can edit them.
The date of the last outbound call attempt recorded for that contact.
# Calls
Source: https://docs.ruleinside.com/conversations
Find a call, see each customer visit, listen to the recording and read what the agent said, did and collected.
Every conversation your agent has is listed under **Calls** in your portal, newest first. Open a call to listen to it,
read the transcript and see what it achieved.
Every member of your team can see the Calls page and every call's page. Nothing on these pages changes your settings
or your data.
## Choosing the agent
At the top of the page, the **Agent** bar shows which agent's calls you are looking at. The Calls page lists the calls
of **one agent at a time**.
* The bar shows the agent's name, with a **Managed** badge if it is your store's default agent. If no agent is chosen
yet it reads **Select agent**.
* Open the list to pick another agent. Your default agent carries the word **managed**, and a tick marks the current
one. The last item, **Manage agents…**, opens the Agents page.
* Your choice is remembered in this browser for 30 days and the page reloads with that agent's calls.
* Any member can switch agent: it only changes what you see.
If your account has no agents yet, or the agent list cannot be loaded, the bar is hidden. When the list cannot be
loaded, the page shows the calls of all your store's agents together.
## The call list
The card is titled **Calls**, with the subtitle "Customer visits for this agent, newest first. One visit can contain
several calls - the call restarts each time the assistant moves the customer to a new page."
The page shows **no total number of calls**: the live call log reports only whether older calls exist, not how many.
For your total, use the **Calls** figure on the [Dashboard](/dashboard).
### Columns
| Column | What it shows |
| - | - |
| **When** | The date and time the call started, in your own time zone and your browser's date format (for example "28 Sept 2026, 13:15"). Click it to open the call. |
| **Customer** | The caller's phone number for phone calls. **Guest** when there is no number, for example on calls from your website widget. |
| **Channel** | **Phone inbound** (the customer called you) or **Phone outbound** (your agent called the customer). Any other call, including website widget calls, shows a dash. |
| **Duration** | The length of the call, for example "59s" or "1m 31s". For a grouped visit, the total of its calls. A dash when the length is not known. |
| **Outcome** | The outcome your agent recorded, for example "Callback requested", "No answer", "Escalated" or "Declined". A dash when none was recorded. |
| **Sentiment** | "Positive", "Neutral" or "Negative". **Not enough speech to judge** when the customer said too little for a judgement. A dash when there is none. |
| **Summary** | A short title or summary of the call, cut to two lines. **Recording deleted at the caller’s request** when the caller objected to being recorded. |
### Loading in two steps
To show the list faster, the page first shows the rows, then fills in **Customer**, **Outcome** and **Sentiment** a
few seconds later. While they load, those cells show a small pulsing grey bar.
If a cell then shows a faint dash with the hover text "Not loaded for this row", that detail could not be loaded this
time (for example because new calls arrived in the meantime). It does not mean nothing was recorded. Refresh the page
to load it again.
## Visits: calls grouped together
When your assistant moves a shopper to another page of your store, the call ends and a new one starts. One visit to
your website can therefore become several short calls. Phone calls from the same number close together are grouped
the same way.
* A row with a small button such as **▸ 2 calls** (or 3, 4 and so on) stands for a whole visit. Hover over it to see
"Show the earlier calls in this visit: the call restarted each time the assistant moved the customer to a new page".
* Click the button to show the earlier calls underneath, indented with an arrow. A "3 calls" visit opens to three rows
in all. Each earlier call has its own link. Click again ("Hide the earlier calls in this visit") to close it.
* The visit row shows the **newest** call's time, link and channel. **Outcome**, **Summary** and **Customer** come
from the newest call that has one. **Sentiment** is the newest real judgement in the visit. **Duration** is the
total of the visit's calls.
Calls belong to the same visit when they come from the same browser session on your website, or from the same caller
number on the phone, **and** no more than **30 minutes** pass between one call and the next.
## Searching
The search box reads **"Search calls: number, summary, outcome, channel…"**.
* It searches only the calls **on the current page**, as you type, and ignores upper and lower case.
* It looks in the date and time, the customer, the channel, the outcome, the sentiment, the call's status and the
summary.
* Outcomes and channels are matched as single words. Search for **callback** or **phone** rather than "Callback
requested" or "Phone outbound".
* Visits are regrouped after a search, so a visit where only some calls match shows a smaller number of calls.
* When nothing matches, the page shows, for example, **"No calls match “shoes”."**
## Newer and Older
At the bottom right, **Newer** and **Older** move through your call history. There are no page numbers.
* A page holds about **50 visits**. It can hold a few more, because a visit is never split across two pages.
* **Newer** is greyed out on the first page; **Older** is greyed out when there is nothing older. When the first page
is also the last, neither button appears.
* Your position is kept in the page's web address, so refreshing, the browser's Back button and a shared link all
land on the same page.
* Switching agent takes you back to the first page.
## Messages on the Calls page
| What you see | What it means | What to do |
| - | - | - |
| **No calls yet.** | This agent has no calls. | If you expected calls, check you have the right agent selected. |
| "This workspace is not linked to a store yet." | Your account is not connected to a store. | Contact support. |
| "We could not load the call log just now. This is a problem on our side, not a sign that there are no calls - please try again in a moment." | The call log could not be read. Your calls are not lost. | Try again in a moment. |
| "Some visits are not shown on this page - there were more than this view can load at once." | There were more visits than the page could load at once, so some are left out of this page. | If a call you need is missing, contact support. |
## A call's page
Click a call's date to open its page, titled **Call**. Press **Back to calls** to return to the list. A call that does
not belong to your store, or that does not exist, shows "Page not found".
The page is made of the cards below, from top to bottom. A card with nothing to show is left out.
### Recording
The **Recording** card has an audio player and the note "Recorded with customer consent. Accessible only to your
team." It appears only when a recording exists for the call. The audio loads when you press play.
If the player shows an error, the recording was deleted at the caller's request, no audio was kept for this call, or
there was a temporary problem: try again in a moment.
### Sentiment over the call
A coloured strip shows how the customer's mood moved through the call, from **Start of call** on the left to **End of
call** on the right. Each segment is one message from the customer: green for **Positive**, grey for **Neutral**, red
for **Negative**. In the middle, **Overall:** shows the overall trend when it is available.
Hover over a segment to see its label, a score from -1 (most negative) to 1 (most positive), and the start of what the
customer said.
RuleInside scores only what the customer said, not the agent, using an AI model, once per call. The strip is not
shown when:
* the customer sent fewer than two messages with real words in them;
* the call opened with a voicemail greeting that RuleInside recognises;
* the caller objected to being recorded;
* the scoring could not be completed at that moment.
This strip is a different measure from the **Sentiment** column in the call list and the **Sentiment** line under
**Data collected**, which are a single judgement of the whole call. They can disagree: for example, a strip
that turns red and then green can sit next to an overall "Neutral".
### Details
| Label | What it shows |
| - | - |
| **When** | The date and time the call started, in your time zone. |
| **Customer** | The caller's phone number; otherwise the customer's email address if the agent has one; otherwise **Guest**. A dash when this could not be read. |
| **Channel** | Widget, Phone inbound, Phone outbound, Whatsapp or Whatsapp outbound. A dash when not known. |
| **Our number** | Your store's phone number used on the call. Shown for phone calls only. |
| **Duration** | The length of the call, for example "18s" or "2m 5s". |
| **Agent** | The name of the agent that handled the call. |
| **Status** | The call's status, for example "Done". |
| **Result** | The automatic verdict on the call: green **Success**, red **Failure**, or grey for anything else (for example "Unknown"). This is the verdict the Dashboard's **Success rate** counts. |
### Summary
A written summary of the call. Not shown when the call has no summary.
### Data collected
One line for each piece of information your agent is set up to collect, with its value. For example **Outcome**,
**Sentiment**, **Lead email**, **Lead interest**, **Lead phone** and **Recording objection**. Any custom fields your
store has appear here too.
A field with no value shows a dash. A yes-or-no field shows "True" or "False".
**Data collected** shows every detail your agent captured, including a phone number or an interest. The Dashboard's
**Leads captured** figure counts only calls that left an email address.
### Evaluation
One block for each evaluation criterion set on your agent (for example "Issue resolved"): its name, a result
(**Success**, **Failure**, or grey for any other value) and an explanation underneath. Not shown when your agent has no
criteria.
### Transcript
The whole conversation, message by message.
* The agent's messages are on the left in grey, labelled **Agent**. The customer's messages are on the right, labelled
**Customer**. Each shows its time into the call, for example "0:11".
* After an agent message's time, a figure such as "· 1.41s" is the **Response latency**: how many seconds the agent
took to respond on that turn.
* **interrupted** (amber) marks a message that was cut off.
* **Actions the agent took**, such as looking up an order or transferring the call, appear as rows with a wrench icon:
the action's name, its type, how long it took, and "ok" (green) or "error" (red). Open one to see **Request** (what
the agent sent) and **Response** (what came back), and the error message if it failed.
* **Knowledge-base lookups** appear as a short row with a book icon, at the moment the agent checked your knowledge
base.
* When there is no transcript, the card reads **No transcript for this call.**
### Notices on a call's page
| What you see | What it means |
| - | - |
| "The caller objected to being recorded, so the recording and transcript of this call were deleted." | The recording, sentiment strip, summary, evaluation, result and transcript are gone. **Data collected** shows only the outcome. The call stays in your list with its outcome and duration. |
| "Call analysis is temporarily unavailable - outcome, sentiment, summary, customer and channel below may be incomplete or blank. The transcript is unaffected." | Part of the call's details could not be read. Reload later. |
| A notice that the live call record could not be loaded, so the transcript, timings and audio below may be incomplete. | The page is showing RuleInside's own copy of the call: the transcript without timings or actions, the summary, and only the outcome and sentiment under **Data collected**. Reload later for the full call. |
| A full-page error, **"Something went wrong"** | The call could not be read at all. Press **Try Again**. |
## When a caller objects to being recorded
If a caller says they do not want to be recorded, the recording and the transcript of that call are deleted. The
call stays in your list, marked **Recording deleted at the caller’s request**, with its outcome and duration.
Your store's privacy notice should tell customers that calls may be recorded. Set the agent's
[recording notice](/managing-agents) so callers hear it.
## Common questions
The Calls page reads the live call log for **the one agent selected** at the top, and groups calls into visits.
The [Dashboard](/dashboard) counts every call RuleInside has recorded for your store, **all agents together**.
They will not match exactly. For your total, use the Dashboard's **Calls** figure.
The Calls list labels only phone calls, as **Phone inbound** or **Phone outbound**. Calls from your website widget
show a dash in **Channel**. A call's own page, and the Dashboard's **Recent calls**, show these calls as **Widget**.
Each time your assistant moves a shopper to another page of your store, one call ends and another starts, so short
calls are normal. They are grouped into one visit: click **▸ 2 calls** (or similar) to see them together.
No phone number is attached to the call, for example because it came from your website.
The list reads your calls live, and a call's summary and outcome appear only once its analysis has finished, a
short time after the call ends. Refresh the page.
Those details are loaded in a second step, which did not return that row this time. Refresh the page.
The list shows one agent at a time. Choose the other agent in the **Agent** bar at the top.
Search for a single word, such as **callback**. Outcomes and channels are matched as one word each. The search also
covers only the calls on the current page; use **Older** to search further back.
They are different measures. The strip follows each of the customer's messages; the column is a single judgement
of the whole call.
No recording exists for that call, the caller objected to being recorded, or there was a temporary problem. Try
again in a moment.
Every member of your team can open a call and play its recording. The note under the player reads "Recorded with
customer consent. Accessible only to your team."
The Dashboard's **Leads captured** counts only calls that left an email address.
# Dashboard
Source: https://docs.ruleinside.com/dashboard
What each figure and chart on your dashboard counts, over which period, and what a notice or error means.
The **Dashboard** is the first page you see when you open your portal. It sums up what your AI voice agent has done:
how many calls it took, how long they lasted, how often it succeeded, the leads it captured, how customers felt, a
14-day chart and your five most recent calls. It is also where notices that need your attention appear first.
## Who can see it
Every member of your team can see the Dashboard. Nothing on it changes your settings or your data: it only shows
figures and links to other pages.
The figures cover **your whole store, all agents together**. Unlike the Calls and Phone numbers pages, the Dashboard
has no **Agent** selector.
## Before your agent is live
Until your subscription is active and your agent is set up, the Dashboard shows a single card instead of the figures.
| What you see | What it means | What to do |
| - | - | - |
| **"Subscribe to activate your AI agent"** | Your account has no subscription yet. | Press **Choose a plan** to go to Billing. |
| **"Complete payment to activate your AI agent"**, with the text "Your voice agent is set up and ready..." | Your agent is ready, but the subscription payment is not confirmed. | Press **Complete payment**. |
| **"Complete payment to activate your AI agent"**, with the text "Choose a plan and complete payment..." | No plan is active yet and no agent has been set up. | Press **Choose a plan**. |
| **"Setting up your AI agent"** | Your subscription is active and our team is setting up your voice agent. | Nothing. The figures appear here as soon as the agent is live. |
Once the subscription is active and your agent is set up, the full Dashboard below appears.
## Notices at the top
Notices appear above the figures in a fixed order: first anything that has stopped, then work waiting for you, then a
suggestion.
**"Outbound calling is paused"** (red) means your account cannot place automated calls until you confirm you have
a lawful basis for calling the people you want contacted. Nothing is sent in the meantime. Press **Review and
confirm** to go to the calling consent confirmation in Settings. This notice cannot be dismissed: it disappears
once you have confirmed.
**"Please re-confirm"** (neutral) means the wording of the calling consent confirmation has changed. Your calls
are still running. Press **Review and confirm** to read the new wording and confirm again. See
[How calling rules protect your customers](/outbound-rules).
**"A refund request is waiting for you"** (or "N refund requests are waiting for you") appears when at least one
shopper has asked your assistant for a refund that you have not decided yet. Nothing is refunded until you
decide. Press **Review refund requests**. See [Refund requests](/refunds).
**"Add a second step to your sign-in"** suggests protecting your login with a second step. Press **Set up
two-step sign-in** to do it, or the **X** to hide the suggestion. Hiding it is remembered only in that browser,
so it can appear again on another computer.
## The four figures
The four figures count every call RuleInside has recorded for your store, for all agents and all time.
### Calls
* **Title:** **"Calls"**, subtitle **"Total voice conversations"**.
* **Big figure:** the total number of calls.
* **Line underneath:** for example **"12 in the last 7 days · 48 in 30 days"**.
* **Period:** the 7 and 30 days are rolling. They count back exactly 7 x 24 hours and 30 x 24 hours from the moment
you opened the page, not from the start of the week or month.
### Talk time
* **Title:** **"Talk time"**, subtitle **"Across analysed calls"**.
* **Big figure:** for example **"84.5 min"**: the total length of your analysed calls in minutes, with one decimal
place.
* **Line underneath:** for example **"Avg call 1m 10s"** or **"Avg call 45s"**: the average length of a call.
* **How it is counted:** only calls with a recorded length between zero and one day count. When no call has a
recorded length, the figure shows a dash.
### Success rate
* **Title:** **"Success rate"**, subtitle **"Resolved by the agent"**.
* **Big figure:** a whole-number percentage.
* **Line underneath:** "No scored calls yet", "Based on 1 scored call" or "Based on 181 scored calls".
* **How it is counted:** after each call, the call is automatically scored as success, failure or unknown. The rate
is the number of successful calls divided by the number of calls scored success or failure. Calls scored unknown,
and calls not scored, are left out of both. When there are no scored calls, the figure shows a dash.
The success rate measures whether the agent achieved the goal of the call. It is not a count of sales.
### Leads captured
* **Title:** **"Leads captured"**, subtitle **"Calls that left an e-mail address"**.
* **Big figure:** the number of calls in which the agent captured an **email address**.
* **Line underneath:** for example **"3 escalated to a human"**: the number of calls whose outcome was "escalated", meaning
the call was handed to a person or transferred.
**Leads captured counts email addresses only.** A call in which the agent captured only a phone number, or only what
the customer was interested in, is not counted here, even though the subtitle says "contact details". You can see
every detail the agent collected on each call's page, under **Data collected** (see [Calls](/conversations)).
## The two charts
### Calls over time
* **Title:** **"Calls over time"**, subtitle **"Daily volume, last 14 days"**.
* **What it shows:** an area chart with one point per day for the last 14 days, including today. Hover over a day to
see its number of **Calls**. Days with no calls show 0. When space is short, not every day label is printed.
* **How days are counted:** calls are grouped by calendar day in **UTC** (Coordinated Universal Time), not in your
local time. A call made just after midnight in your time zone can therefore appear on the previous day.
### Sentiment
* **Title:** **"Sentiment"**, subtitle "No analysed calls yet", "Across 1 analysed call" or "Across N analysed calls".
* **What it shows:** three bars, **"Positive"**, **"Neutral"** and **"Negative"**: how many calls the agent judged in
each mood after the call.
* **How it is counted:** the subtitle number is the sum of the three bars. Calls with no sentiment are left out.
This is the agent's overall judgement of each call. It is a different measure from the **Sentiment over the call**
strip on a call's own page, which follows the customer's mood message by message.
## Recent calls
Title **"Recent calls"**, subtitle **"The latest conversations"**. It lists the **5** most recent calls of your store,
all agents together, newest first. Press **View all** to open the [Calls](/conversations) page, or a call's date to
open that call.
It has the same columns as the Calls page (**When**, **Customer**, **Channel**, **Duration**, **Outcome**,
**Sentiment**, **Summary**) and a search box, with a few differences:
| Column | On the Dashboard |
| - | - |
| **Customer** | The customer's email address, if the agent has one for them. Otherwise **"Guest"**. |
| **Channel** | Widget, Phone inbound, Phone outbound, Whatsapp or Whatsapp outbound. |
| **Duration** | The analysed length of the call; if that is missing, the time between the call's start and end. |
The search box filters only these five rows. Calls are not grouped into visits here, and there are no **Newer** or
**Older** buttons.
## Errors and notices
| What you see | What it means | What to do |
| - | - | - |
| A full-page error, "500", **"Something went wrong"** and a **Try Again** button | The figures could not be read. The Dashboard never shows a figure it could not read as zero, and never guesses your subscription state. | Press **Try Again**. If it keeps happening, contact support. |
| A red-bordered notice: "Call analysis for the recent calls is temporarily unavailable..." | Only the details of the five recent calls could not be read: their outcome, sentiment, summary, customer and channel are blank and their durations may show as 0. | Nothing to fix: the four figures and the two charts come from a separate count and are correct. Reload later. |
| A full-page loader | The page is loading. | Wait a moment. |
## Common questions
They count different things. The Dashboard counts every call RuleInside has recorded for your store, **all
agents together**, over all time. The [Calls](/conversations) page reads the live call log for **the one agent
selected** at the top, and groups calls into customer visits. It also shows no total. The two will not match
exactly. For your total number of calls, use the Dashboard's **Calls** figure.
**Leads captured** counts the calls that left an **email address**. A call in which the agent captured only a
phone number, or only what the customer was interested in, is not counted. Open a call and look under **Data
collected** to see everything the agent captured.
No. It is the last 7 x 24 hours, counted back from the moment you opened the page. The same applies to 30 days.
The chart groups calls by UTC calendar day, not by your local day. A call made shortly after midnight in your time
zone can appear on the previous day.
Only calls scored success or failure count. Calls scored unknown, and calls that were not scored, are left out.
The line under the figure tells you how many calls it is based on.
The calls whose outcome was "escalated": the agent handed the customer to a person or transferred the call.
A figure could not be read at that moment, and the Dashboard does not show zeros in its place. Press **Try
Again**. If the error persists, contact support.
No. It disappears once you confirm the calling consent in Settings, using **Review and confirm**.
Either you hid it with the **X** in this browser, or you have set up a second sign-in step.
Not on the Dashboard: it always covers all your agents together. To look at one agent's calls, use the
[Calls](/conversations) page and choose the agent at the top.
# FAQ
Source: https://docs.ruleinside.com/faq
Answers to the questions store owners ask most, with a link to the page that has the full detail.
Short answers to the questions store owners ask most. Each answer links to the page with the full detail. For the
words used here, see the [Glossary](/glossary).
## Getting started
WooCommerce and FluentCart stores. A WooCommerce store owner can connect the store from the portal; a FluentCart
store is connected by our team. See [Connections](/connections).
No. Our team sets up and manages your agents. You change their voice, greeting and instructions from the portal,
in plain language. See [Managing your agents](/managing-agents).
Follow the checklist on [Your first agent](/your-first-agent): look at the Dashboard, read and try your agent,
check your website assistant and phone number, confirm calling consent and invite your team.
The website assistant runs inside its own frame on your page, separate from your store's code.
On a WordPress store with the RuleInside plugin, you do not paste the widget snippet into your site: you paste
only your widget's **embed id** into the plugin's **Widget ID** field (WordPress admin, menu **RuleInside**).
Anyone on your team can download the plugin from **Settings**, **Plugin**. A licence key, pasted into the
plugin's **License Key** field, gives your store automatic plugin updates. See
[Website assistant](/website-assistant) and [WordPress plugin](/plugin).
## Your agent
Two things decide it. Only an **owner** can change an agent; a **member** can read everything but change nothing.
Then the agent's **edit level** (**View only**, **Name, voice & greeting**, **Instructions** or **Full control**)
decides which settings an owner can change. Your main agent keeps the level our team set; agents you create
start at **Full control**. See [Who can do what](/roles) and [Edit levels](/agent-settings#edit-levels).
Yes, if you are an owner and the agent is at **Name, voice & greeting** or above. Edit the **First message** card
and click **Save changes**. Keep a mention that the caller is talking to an AI assistant and that the call may be
recorded. See [First message](/agent-settings#first-message).
The language you set for it in the agent's settings. Each agent has one **Default language**, chosen in the
**Language** panel of the agent editor (this needs the **Instructions** level). If you serve shoppers in another
language too, you can create a second agent set to that language. See [Language](/agent-settings#language) and
[Creating an agent](/managing-agents#creating-an-agent).
Yes, for a customer who is logged in to your store, and only their own orders. A customer who is not logged in
gets an order's details only by giving its order number together with the billing email and postcode on that
order.
No. It records the refund request, and you decide it under **Refund requests**. Approving only records your
decision: you refund the order in your store and tell the shopper yourself. See [Refund requests](/refunds).
Yes. Owners click **Preview** in the agent editor to talk to it by voice or text. Preview treats you as a guest,
so it gives order details only for an order number with its billing email and postcode. For outbound calls, a campaign can call your own **Test numbers** first; that
is a real, billable call. See [Testing an agent with Preview](/managing-agents#testing-an-agent-with-preview) and
[The campaign page](/campaign-results#send-test-call).
## Calls and messages
Yes. Recordings and transcripts are kept in your portal for your team. If a caller says they do not want to be
recorded, that call's recording and transcript are deleted. Your privacy notice should tell customers that calls
may be recorded. See [Calls](/conversations).
Every member of your team can open a call under **Calls** and play its recording. There is no recording when the
caller objected to being recorded or no audio was kept. See [Calls](/conversations).
On phone calls, when our team has switched **Transfer to a person** on for your account. You can then add up to
5 of your numbers, and say when the agent should hand over. It does not work on website or WhatsApp
conversations. See [Transfer to a person](/agent-settings#transfer-to-a-person).
No. WhatsApp only lets a business start a conversation with a template Meta has approved. You create templates in
Meta's WhatsApp Manager, register them under **WhatsApp templates**, and send them one at a time from
**WhatsApp send** or with a campaign. See [WhatsApp](/whatsapp).
## Customers and consent
Yes, with campaigns, within the rules that protect your customers: consent, do-not-call, calling hours and
weekly limits. See [Campaigns](/campaigns) and [How calling rules protect your customers](/outbound-rules).
No. Every call and message is checked just before it goes out. Someone who opted out on a channel is never
contacted on it again, and if a check cannot be completed, the person is not contacted. See
[How calling rules protect your customers](/outbound-rules).
A plain import adds people with no consent. Record a lawful basis with **Import with consent**, or with
**Consent** on each contact. Then check that an owner has confirmed **Settings**, **Calling consent**: until
then, nothing is sent. Finally, check your calling hours. See [Contacts](/contacts) and
[Calling consent and calling hours](/calling-hours).
On **Contacts**, click **Opt out** on that person. Any member can do it. It stops both calls and WhatsApp and
cancels anything already scheduled for them. You cannot undo it: only the customer can opt back in, and only on
WhatsApp, by replying START. A customer can also stop WhatsApp messages and calls themselves by replying STOP on WhatsApp. See
[Opting a contact out](/contacts#opting-a-contact-out).
No. Only owners can create, test, launch, pause, resume or cancel a campaign. Members can open campaigns, read
the results and download them. See [Campaigns](/campaigns).
Open the campaign. **Pause** (phone campaigns) stops new calls and keeps everyone's place, and **Resume** carries
on later. **Cancel** ends the campaign for good. Calls already under way cannot be recalled. Pausing and
cancelling work even when your account is not active. See
[The campaign page and its results](/campaign-results#the-buttons).
## Account and billing
On Lite, Starter, Growth and Scale, your assistant keeps working and the extra minutes are billed as extra
usage. On a pay-as-you-go bundle, you are told at 80% and 100%, and the same bundle is bought again with your
saved card, at most twice in a billing period. After that, outbound calls and messages can be stopped for the
rest of the period, while inbound calls keep working. The **Minutes** card on the **Billing** page shows how many
minutes you have left. See [Plans, limits and what happens at the limit](/limits).
An owner clicks **Cancel plan** on the **Billing** page. You keep full access until the end of the term you
paid for, and you can undo it until then with **Don't cancel**. Unused minutes are not refunded. See
[Cancelling your plan](/billing#cancelling-your-plan).
Every account is isolated. Your team sees only your own calls, contacts and customer data, never another
business's.
Email us at [info@ruleinside.com](mailto:info@ruleinside.com). If it is about a call, include the agent's
**Agent ID** and the date and time of the call.
# Glossary
Source: https://docs.ruleinside.com/glossary
The words you meet in the RuleInside portal and in these docs, explained in plain terms, with a link to the page that covers each one.
The words you meet in your portal and in these docs, in alphabetical order. Each entry links to the page with the
full detail.
## A
**Additional instructions**: Your own guidance for an agent, written in plain language and added to its instructions in a section of its own.
The rest of the instructions stay managed by RuleInside. It needs the **Instructions** edit level. See
[System prompt](/agent-settings#system-prompt).
**Agent**: One AI assistant, with its own voice, greeting, instructions, knowledge and tools. Your account can have several
agents. See [Managing your agents](/managing-agents).
**API key**: A long password that lets your own software talk to RuleInside, with only the permissions you tick when you create
it. It is shown once and never again. See [API keys](/api-keys).
**Automation**: A standing rule that places calls on its own when contacts match a condition, for example no activity for 30
days. There is nothing to launch: a rule that is switched on runs by itself. See [Automations](/automations).
## C
**Calling consent**: The confirmation an owner gives under **Settings**, **Calling consent**, that your store has a lawful basis to have
an AI agent contact the people it asks RuleInside to contact. Until it is confirmed, no calls and no WhatsApp messages
are sent. See [Calling consent and calling hours](/calling-hours).
**Calling hours and days**: When your agent may call, in each customer's own local time. By default, 08:00 to 21:00, every day. A campaign can
use its own. See [Calling hours and days](/calling-hours#the-calling-hours-and-days-card).
**Campaign**: A one-off outreach to a group of your contacts, by phone call or by WhatsApp template message. See
[Campaigns](/campaigns).
**Concurrent calls**: How many of a campaign's calls may ring or be under way at the same time. A number above your plan's maximum is
lowered to it. See [Plans and limits](/limits#campaigns).
**Consent**: On a contact, a recorded lawful basis to contact that person on **one** channel. Calls and WhatsApp are recorded
separately, and one never implies the other. See [Consent badges](/contacts#consent-badges).
**Contact**: A person in your store's list for outbound calls and WhatsApp messages. The phone number is the contact's key:
consent and the do-not-call list are tied to it, and it cannot be edited. See [Contacts](/contacts).
**Custom field**: An extra column on your contacts, such as `order_number`, added by import or with **Add column**. Your agent can use
it on an outbound call as a variable. See [Custom fields](/importing-contacts#custom-fields).
## D
**Do-not-call list**: Your store's list of numbers that must never be contacted, kept separately for calls and for WhatsApp (where it is
called the do-not-contact list). Opt-outs are added to it automatically. There is no page to view or edit it. See
[Your do-not-call list](/contacts#your-do-not-call-list).
**Draft**: A campaign that is saved but not launched. Nothing is dialled or sent until an owner presses **Launch campaign**,
and only a draft can be edited. See [Editing a draft](/campaigns#editing-a-draft).
## E
**Edit level**: What an owner can change on one agent: **View only**, **Name, voice & greeting**, **Instructions** or **Full
control**. Each level includes the ones before it. See [Edit levels](/agent-settings#edit-levels).
**Embed id**: The code inside your website widget's snippet that says which widget to load. The WordPress plugin calls it
**Widget ID**. See [Website assistant](/website-assistant#install).
## F
**First message**: What the agent says first: its greeting. It should tell callers they are speaking with an AI assistant and that the
call may be recorded. The **Recording notice** setting next to it decides where the recording sentence is said. See
[First message](/agent-settings#first-message).
## I
**Inbound agent and outbound agent**: On **Phone numbers**, the agent that answers calls to a number, and the agent that places outbound calls from it.
A dash means none. See [Assign agents](/phone-numbers#assign-agents).
**Inbound connection**: A web address and a secret that let your other systems, such as a CRM or a booking tool, send contacts and events
into RuleInside. Incoming data never places a call by itself and never gives consent. See
[Inbound connections](/connections#inbound-connections).
## K
**Knowledge base**: The documents an agent answers from: web pages, files and text, such as your policies and FAQs. See
[Knowledge](/knowledge).
## L
**Lawful basis**: The reason you are allowed to contact a person, for example "Web / checkout opt-in" or "Existing customer
relationship". You record it on the contact; RuleInside does not check that it is true. See
[Lawful bases](/contacts#lawful-bases).
## M
**Main agent and the Managed badge**: Your main agent is the one our team set up for you. It carries a **Managed** badge and cannot be deleted. The badge
does not lock anything by itself: the agent's edit level decides what you can change. See
[The agent selector](/managing-agents#the-agent-selector).
**Member**: A role in your organization. A member can see every page and invite colleagues as members, but cannot change how
your assistant works. A member can still opt a contact out. See [Who can do what](/roles).
## O
**Opt-out**: A person's request not to be contacted. **Opt out** on a contact, open to every member, stops both calls and
WhatsApp. You cannot reverse an opt-out: only the customer can, on WhatsApp, by replying START. See
[Opting a contact out](/contacts#opting-a-contact-out).
**Owner and primary owner**: An owner can see and change almost everything, including billing and members' roles. One owner per organization is
the **primary owner**, who alone can rename the organization, change its logo and transfer ownership. See
[Who can do what](/roles).
## P
**Personalization variable**: A value your storefront passes in when a call starts, such as the customer's first name. You use it in the first
message or the instructions by writing its name in double curly braces, for example `{{first_name}}`. See
[Personalization variables](/agent-settings#personalization-variables).
**Plugin**: The RuleInside plugin for WordPress. You download it from **Settings**, **Plugin**, paste your widget's embed id
into its **Widget ID** field, and paste a **licence key** into its **License Key** field so it receives updates. See
[WordPress plugin](/plugin).
**Preview**: A test conversation with an agent in your browser, by voice or text, for owners. It uses the agent's saved settings
and treats you as a guest. See [Testing an agent with Preview](/managing-agents#testing-an-agent-with-preview).
**Prompt mode and Auto mode**: How an agent uses one of your documents. A **Prompt** document is always given to the agent; an **Auto** document
is looked up when it is relevant to the question. See
[How the agent uses a document](/knowledge#how-the-agent-uses-a-document).
## R
**RAG**: Short for Retrieval-Augmented Generation. It lets the agent pull only the relevant parts of large documents into
each answer, instead of loading whole documents. You switch it on with **Configure RAG** on the Knowledge page. See
[Configure RAG](/knowledge#configure-rag).
**Refund request**: A refund a recognised shopper asked your assistant for. You approve or decline it under **Refund requests**.
Nothing is refunded automatically: you refund the order in your store. See [Refund requests](/refunds).
**Retry**: When a campaign call is not answered, the campaign calls the person again on its own: at most 3 attempts in all, at
most 2 in one day, at least 1 hour apart. See [Retries](/campaign-results#retries).
## S
**STOP and START**: Replies a customer can send on WhatsApp. **STOP** opts them out of your WhatsApp messages and your calls;
**START** opts them back in to WhatsApp messages only (calls stay stopped). See [STOP and START](/whatsapp#stop-and-start).
**System prompt**: The agent's full written instructions. You can edit it as a whole only at **Full control**, and RuleInside's own
sections must stay in it. At lower levels you add **Additional instructions** instead. See
[System prompt](/agent-settings#system-prompt).
## T
**Template category**: The category Meta gives a WhatsApp template: **Marketing**, **Utility** or **Authentication**. A contact receives
at most 4 Marketing messages in any 7 days, and, except for immediate follow-ups sent within 24 hours of the
shopper's chat, Marketing messages wait for the person's calling hours. See
[How the category affects sending](/whatsapp-templates#how-the-category-affects-sending).
**Test number**: One of your own phone numbers, added to a phone campaign so you can hear the agent with **Send test call** before
launching. A test call is a real, billable call. See [Send test call](/campaign-results#send-test-call).
**Tool**: Something an agent can do during a conversation, such as searching your products or looking up an order. Our store
tools are marked **Managed**. When tools are switched on for your account, an owner can add their own on an agent at
**Full control**. See [Tools](/tools).
**Transfer to a person**: On phone calls, the agent hands the caller over to one of your numbers when the situation calls for it. It works
only when RuleInside has switched it on for your account, or has set a transfer up for you. See
[Transfer to a person](/agent-settings#transfer-to-a-person).
## W
**Webhook**: A signed message RuleInside sends to an address on your own system when something happens, such as a call ending or
a contact opting out. See [Webhooks](/webhooks).
**Website assistant (widget)**: The voice and chat box on your website, through which shoppers talk or type to one of your agents. Each widget
belongs to one agent. See [Website assistant](/website-assistant).
**WhatsApp template**: A message Meta has approved, which WhatsApp requires to start a conversation. You create it in Meta's WhatsApp
Manager and register it in RuleInside with a purpose, such as **Re-engagement (campaigns)** or **Manual send**. See
[WhatsApp templates](/whatsapp-templates).
# How it works
Source: https://docs.ruleinside.com/how-it-works
What happens on a call, where shoppers meet your assistant, what RuleInside manages and what you control.
This page explains, in plain terms, what your AI assistant does with a shopper, what our team manages for you, what
you can change yourself, and the rules that protect your customers. It takes about five minutes to read.
## What happens on a call
The shopper opens the assistant on your website, calls your phone number, or messages your WhatsApp number. The
assistant greets them with its **first message**, which should tell them they are speaking with an AI assistant
and that the call may be recorded (see [First message](/agent-settings#first-message)).
It follows its instructions and uses the knowledge it has been given, such as your policies, FAQs and product
details. See [Knowledge](/knowledge).
When it needs a fact, the assistant uses a **tool**: it searches your products, looks up an order for a customer
who is logged in to your store (or who gives the order number with the billing email and postcode on it), or
checks a coupon code. When tools are switched on for your account, your own tools can be added too. See
[Tools](/tools).
When the conversation ends, it appears under [Calls](/conversations), with the recording when one exists, the
transcript, a summary, the outcome, how the customer felt and the details the assistant collected. The actions
the assistant took, such as looking up an order, are shown in the transcript.
If a caller says they do not want to be recorded, the recording and transcript of that call are deleted. The call
stays in your list, marked **Recording deleted at the caller’s request**.
## Where shoppers meet it
| Channel | How shoppers reach it | Where you set it up |
| - | - | - |
| **Your website** | A voice and chat box on your store's pages. Shoppers can talk, type, or both. | [Website assistant](/website-assistant), and on WordPress the [plugin](/plugin) |
| **The phone** | Calls to a phone number you have brought in from Twilio are answered by the agent you assign to it. The same number can also place outbound calls. | [Phone numbers](/phone-numbers) |
| **WhatsApp** | Shoppers message your WhatsApp Business number. On a number our team connected for you, the agent assigned to it answers them. You can also send approved template messages to customers who agreed to receive them. | [WhatsApp](/whatsapp) |
Your assistant can also **call your customers**, with a [campaign](/campaigns) you launch or an
[automation](/automations) you switch on, always within the rules described below.
## What it can do
Searches your catalogue by meaning, not only by exact words, and points the shopper to the right product.
For a customer logged in to your store, or one who gives the order number with the billing email and postcode
on it: the status of their own orders, and the tracking number when your store records one.
Tells the shopper whether a coupon code is valid.
Uses the knowledge you give it, such as your policies and FAQs. See [Knowledge](/knowledge).
Records the request for your team. A person approves every refund; nothing is refunded automatically. See
[Refunds](/refunds).
Campaigns call the people you choose, only within the rules that protect them. See [Campaigns](/campaigns).
When our team has set them up for your account, it can also hand a phone caller over to one of your numbers
([Transfer to a person](/agent-settings#transfer-to-a-person)), and book a call-back a customer asks for on a
campaign call ([Call-back states](/campaign-results#call-back-states)).
## What it will not do
* **It never refunds money.** It records the request, and you decide. Approving in the portal only records your
decision: you refund the order in your store and tell the shopper yourself. See [Refund requests](/refunds).
* **It does not give order details to just anyone.** Only a customer logged in to your store, or one who gives the
order number together with the billing email and postcode on that order, hears about it. A refund request
can be recorded only for a shopper your store has recognised.
* **It never contacts someone who asked to stop**, has no recorded consent for that channel, or is outside their
calling hours. See [How calling rules protect your customers](/outbound-rules).
* **It never starts a WhatsApp conversation with free text.** WhatsApp only allows a business to start with a
template Meta has approved. See [WhatsApp templates](/whatsapp-templates).
* **It cannot transfer a website or WhatsApp conversation to a person.** Transfer to a person works on phone calls
only.
## What RuleInside manages and what you control
Our team builds your main agent and keeps the parts that make it work in our hands, so an edit cannot break them:
* the store tools (product search, order look-up, coupon check), which always stay on the agent;
* the knowledge documents we add, marked **Managed**;
* our sections of the agent's instructions, and the language model it uses;
* what happens when a call reaches voicemail (our policy is that the agent hangs up and leaves no message).
What **you** can change on an agent depends on two things: your role, and the agent's **edit level**.
* **Your role.** Only an **owner** can change agents. A **member** can open everything and read every setting, but
cannot change how the assistant works. See [Who can do what](/roles).
* **The edit level.** Every agent has one of four levels, shown as a badge in the agent editor. Each level includes
the ones above it:
| Edit level | In plain terms, you can change |
| - | - |
| **View only** | Nothing. Contact us for changes. |
| **Name, voice & greeting** | The agent's name, its voice, its greeting and its personalization variables. |
| **Instructions** | Also its language, how it sounds, how long it waits before replying, how long calls last, transfers to a person, the recording notice, and your own **Additional instructions**. |
| **Full control** | Also the whole system prompt, the speech engine and sound format, the knowledge documents, the tools and the built-in system tools. |
Your main agent, marked **Managed**, keeps the level our team set for it. Agents you create yourself always start at
**Full control**. The full list of settings for each level is on [Edit levels](/agent-settings#edit-levels).
Outside the agent itself, owners also run the website widget, phone numbers, WhatsApp, contacts, campaigns,
automations, refund decisions, connections and billing.
## The rules that protect your customers
Every outbound call and every outbound WhatsApp message is checked, person by person, just before it goes out. You do not
switch these checks on: they apply to every account.
1. Your store has confirmed, under **Settings**, **Calling consent**, that it may contact its customers. Until an
owner confirms, nothing outbound is sent.
2. The person has not opted out on that channel.
3. Their number is not on your do-not-call list.
4. A lawful basis is recorded for that channel. Consent to calls and consent to WhatsApp are separate.
5. For calls, it is within your calling hours, in the person's own time zone (8:00 to 21:00 unless you set other
hours).
6. The person has not been contacted too often, and your account is within its daily limit.
If a check cannot be completed, the person is not contacted. Read the details on
[How calling rules protect your customers](/outbound-rules).
## Where it runs
RuleInside works with **WooCommerce** and **FluentCart** stores. The website assistant runs inside its own frame on
your page, separate from your store's code. A WooCommerce store owner can connect the store from the portal; a
FluentCart store is connected by our team. See [Connections](/connections).
## Next steps
Work through the first-day checklist.
Every setting in the agent editor.
The checks before every call and message.
The words used in the portal, in plain terms.
# Importing contacts
Source: https://docs.ruleinside.com/importing-contacts
Bring your customer list in from a CSV, with or without a lawful basis, and use your own columns in your agent.
You bring customers into **Contacts** from a CSV file. There are two ways to do it, and the difference is whether
the people in the file can be called straight away.
| | **Import contacts** | **Import with consent** |
| - | - | - |
| What it does | Adds the people in the file, or updates the ones you already have | Adds or updates them **and** records the lawful basis to call you choose, for everyone in the file |
| Badge they arrive with | **Calls: no consent** | **Calls: callable** |
| Can they be called? | No, not until you record a lawful basis | Yes, subject to the other checks (opt-outs, do-not-call, calling hours, limits) |
| Phone numbers | Stored as written in the file | Converted to international format; a number that cannot be converted is skipped |
| Who can use it | Owners | Owners |
Both buttons are at the top of the **Contacts** page. They are greyed out until a store is linked to your
workspace. **Import with consent** is also greyed out if the list of lawful bases could not be loaded; refresh the
page and try again.
**Import with consent** always records consent to **call**. There is no bulk import that records WhatsApp
consent. You record WhatsApp consent one contact at a time with the **Consent** button. See
[Contacts](/contacts#recording-consent-for-one-contact).
Members who are not owners see this note instead of the buttons: "Only account owners can import contacts or change
consent - ask an owner of your team." See [Who can do what](/roles).
## Before you import
Write every phone number in international format: a `+`, the country code, then the number, with **no spaces**.
For example `+15551234567`. Numbers written this way can be called, granted consent and opted out without
surprises. **Import with consent** skips any number without a country code.
* **The file must be a CSV.** The upload accepts `.csv` files only. If your list is in Excel, save it as CSV first.
* **At most 1,000 contacts per import.** Split a longer list into several files.
* **One email per contact.** An email address can belong to only one contact in your store. A row whose email
already belongs to a different contact is skipped.
## Import step by step
On **Contacts**, click **Import contacts** or **Import with consent**. The dialog asks you to "Paste or upload a
CSV with a phone column (plus optional name and email). Phone numbers should be in international format (e.g.
+15551234567)."
Either choose a file under **CSV file (optional)**, or paste the CSV text into **Or paste CSV**. A file you
choose is loaded into the paste box, so you can check it before importing. Need a starting point? Click
**Download CSV template**.
As soon as there is text in the box, the dialog shows how many contacts are ready, how many rows will be
skipped because they have no phone, which custom fields it found, and which column names it will ignore.
Pick the basis under **Lawful basis to call**. For two of the bases you must also tick **I confirm the statement
above.** See [Import with consent](#import-with-consent).
Click **Import** (or **Import + record consent**). The button shows "Importing…" while it runs, and a message
tells you the result.
## Every field in the dialog
| Field | What it does | Values and limits |
| - | - | - |
| **CSV file (optional)** | Upload a file instead of pasting. Its text appears in the paste box. | `.csv` only. Excel files are not accepted here. |
| **Download CSV template** | Downloads `contacts-import-template.csv`, a ready-made file with two example rows. | Columns: `phone,name,email,city,order_number,language`. Everything after `email` is an example of a custom field. |
| **Or paste CSV** | Paste CSV text directly. The box shows an example: a header line `phone,name,email` and one row. | Any CSV text. |
| **Lawful basis to call** | Import with consent only. The basis that applies to **every** contact in the file. Starts as "Choose a basis…". | One of the calling bases (see below). |
| **I confirm the statement above.** | Import with consent only. Appears under a fixed statement for **Existing customer relationship** and **Purchased marketing list**. | Must be ticked to import. Changing the basis clears the tick, so a tick you gave for one statement never covers another. |
**Buttons:**
* **Cancel** closes the dialog. Nothing is imported.
* **Import** (or **Import + record consent**) runs the import. It stays greyed out until at least one contact is
ready, the file is within the 1,000 limit, and, for Import with consent, a basis is chosen and any required
statement is ticked.
## How your CSV is read
**The header row.** If the first row names a phone column, it is treated as a header. Columns can be in any order.
Names are matched ignoring upper and lower case, and spaces or hyphens count as underscores, so "Phone Number",
"E-mail" and "First Name" all work.
| Column | Header names recognised |
| - | - |
| Phone (required) | phone, phone\_number, phone\_no, number, telephone, tel, mobile, mobile\_number, mobile\_phone, cell, cellphone, cell\_phone, contact\_number, msisdn |
| Name (optional) | name, full\_name, fullname, first\_name, firstname, contact, contact\_name, customer, customer\_name |
| Email (optional) | email, email\_address, emailaddress, e\_mail, mail |
**No header row.** If the first row does not name a phone column, the columns are read as phone, name, email, in
that order. No custom fields are imported, because there is nothing to name them.
**Other details:**
* Values in quotes can contain commas, line breaks and doubled quotes.
* Fully blank lines are ignored.
* A row with no phone value is skipped and counted.
* If the same contact is in the file twice, it counts once when you import with consent.
### A sample file
```csv theme={null}
phone,name,email,order_number
+15551234567,Jane,name@example.com,1042
```
This imports one contact with a custom field called `order_number`.
## Custom fields
Every column that is not phone, name or email becomes a **custom field**, named after its header. Your agent can use
it on an outbound call as a variable: a column called `order_number` is available as `{{order_number}}`. The
preview lists the custom fields it found, and reminds you that re-importing a contact replaces its custom fields.
**Naming rules for a custom field:**
* letters, numbers and underscores only;
* must start with a letter or an underscore;
* at most 64 characters;
* must not start with `system__` or `_ri_`;
* must not be `customer_name` or a small number of other reserved names.
A column with a name that breaks these rules is listed in red as "Ignored (invalid field names)" and is not
imported.
**Other limits:**
* A value can be at most 512 characters. A longer value is rejected, not shortened.
* A contact can hold at most 30 custom fields.
* Fields that are rejected are counted in the result message ("custom field(s) ignored").
* A second column that is just another name for phone, name or email (for example a "Mobile" column next to "Phone")
is ignored, so phone numbers never end up in a custom field.
* If two headers have the same name, the first one is used.
* Blank cells are skipped.
New field names from an import are added to your workspace's columns automatically, up to 30 columns in total.
## What happens to contacts you already have
A contact is matched by phone number. Importing a number you already have **updates** that contact rather than
creating a second one.
* **Name and email** are written only when the file has a value. A blank cell never erases a name or email you
already have.
* **Custom fields:** if the row has at least one custom value, the contact's custom fields are **replaced** by the
row's set. If the row has no custom values, the stored ones are left alone.
* **Opt-outs and the do-not-call list are never touched** by an import. Someone who opted out stays opted out, even if
you import their number again.
* A plain import never changes consent. An import with consent records the basis you chose, replacing any earlier
calling basis for those numbers.
## Import with consent
Use this when you already have a lawful basis to call everyone in the file. The dialog reminds you: "Recording
consent lets your agents call these contacts. You are responsible for having a lawful basis for each person."
**The bases you can choose under Lawful basis to call:**
| Basis | Statement to confirm? |
| - | - |
| Web / checkout opt-in | No |
| Explicit written consent | No |
| Verbal consent (recorded call) | No |
| Existing customer relationship | Yes |
| Purchased marketing list | Yes |
The meaning of each basis, and the full text of the two statements, are on [Contacts](/contacts#lawful-bases).
**What happens when you import with consent:**
* Each phone number is converted to international format. A leading `00` counts as `+`, and spaces, dashes and
brackets are removed.
* A number that cannot be converted, including a number with no country code, is skipped and counted as "invalid
number".
* Each contact is added or updated, and the basis is recorded with today's date.
* One record of the import is kept: who did it, which basis, how many contacts and, for the two bases with a
statement, the exact statement you confirmed.
* If no number in the file can be used, the import is refused.
RuleInside does not check that the basis you choose is true. You are responsible for it.
## Result messages
| Message | Meaning |
| - | - |
| "Imported 12 contact(s)." | Plain import finished. It may add how many rows were skipped and how many custom fields were ignored. |
| "Consent recorded for 12 contact(s)." | Import with consent finished. It may add "skipped (invalid number)" and how many custom fields were ignored. |
| A red note in the preview saying the file is over the limit and to split it into smaller files | The file has more than 1,000 contacts and cannot be imported as it is. |
| "Could not import the contacts." | The plain import failed. Nothing to undo; try again. |
| "Could not record consent." | The import with consent failed. |
## Showing, hiding and adding columns
The **Columns** button on the Contacts page opens a panel with two halves. When some columns are hidden, the button
shows how many (for example "Columns -2").
### Show columns (your own view)
Anyone in the workspace can use this half. It appears once you have contacts.
* Tick or untick a column to show or hide it. **Phone**, **Consent** and **Actions** are marked "Always shown" and
cannot be hidden.
* Drag a row to change the order. **Actions** always stays last. Dragging works with a mouse; it does not work on
touch screens or with the keyboard.
* **Reset to default** clears your choices and your order.
* By default, the built-in columns are shown, and a custom column is shown only when at least one contact on the
current page has a value in it. Your own ticks override this either way.
These choices apply **to this browser only**. A colleague, or you on another device, sees their own view.
### Workspace fields (owners only)
The bottom half changes the columns for everyone: "Changes here affect everyone in this workspace, not just your
view." It is there even before your first import.
**Add column** opens the **Add a column** dialog.
* Type a name under **Column name** (for example `order_number`). The same naming rules as custom fields apply.
* If the name is not allowed you see "Use only letters, numbers, and underscores (and not a reserved name)."; if it
is already a column, "That column already exists."
* Click **Add column**. You see "Column added." and the new column is switched on in your browser.
* If the name is refused, for example because it is reserved or you already have 30 columns, you see "That name is
not a valid column and was ignored."
The new column is a variable you can use in your agent's prompt on outbound calls, for example `{{order_number}}`.
You fill in its value per contact with **Edit** on the Contacts page, or with a new import.
**Manage fields** (with the number of fields, for example "Manage fields (3)") opens the **Columns** dialog, listing every workspace field.
* Click **Remove** next to a field, then **Save**. Nothing changes until you save. You see "Columns updated."
* Removing a column **does not delete** the values stored on your contacts. Adding the column back shows them again.
While a column is removed, its values are **not sent to your agent** on outbound campaign calls. If your agent's
prompt uses that variable, add the column back before you launch a campaign.
A workspace can have at most 30 columns.
## Common questions
Not on the Contacts page. Save it as CSV first. (The recipient list you upload when you create a campaign does
accept Excel files. See [Campaigns](/campaigns).)
Only a phone column. Name and email are optional. Any other column becomes a custom field. Download the template
from the dialog to start from a working file.
1,000 per import. Split a longer list into several files and import them one after another.
With **Import contacts**: the row had no phone number, or its email already belongs to another contact. With
**Import with consent**: also any number that could not be put in international format, usually because it has
no country code. These are counted as "invalid number".
A plain import adds people as **Calls: no consent**. Record a lawful basis for them, with **Import with
consent** or the **Consent** button on each contact. If the badges already say **Calls: callable**, check
[Settings, Calling consent](/calling-hours): nothing is sent until an owner has confirmed it.
A row with at least one custom value replaces the contact's whole set of custom fields. If you want to keep a
field, include its column and value in the new file. A row with no custom values leaves them alone.
No. An import never touches opt-outs or your do-not-call list. See [Contacts](/contacts#opting-a-contact-out).
No. **Import with consent** records consent to call. Record WhatsApp consent per contact with the **Consent**
button.
It switches on automatically in the browser you added it from. In another browser, an empty custom column starts
hidden: tick it in **Columns**.
No. The values stay stored on your contacts and come back when you add the column again. While the column is
removed, those values are not sent to your agent on outbound calls.
No. Importing, recording consent and changing workspace fields are for owners. Any member can still choose their
own visible columns, and can opt a contact out.
# Welcome to RuleInside
Source: https://docs.ruleinside.com/index
An AI voice and WhatsApp assistant for your WooCommerce or FluentCart store, set up by our team and run by you from your portal.
**RuleInside gives your online store an AI assistant that talks to your shoppers on your website, on your phone line
and on WhatsApp, and can call your customers, within the rules that protect them.**
Our team sets your assistant up and manages the parts that keep it working. You run it from your portal: change
what it says, read every conversation, reach your customers and decide refund requests, without touching code.
## What do you want to do?
Find your task on the left, then open the page on the right.
| I want to... | Page |
| - | - |
| Get going on my first day | [Your first agent](/your-first-agent) |
| Understand what the assistant does on a call | [How it works](/how-it-works) |
| Change what the assistant says first (the greeting) | [Agent settings: First message](/agent-settings#first-message) |
| Change its voice, language or how quickly it replies | [Agent settings](/agent-settings) |
| Add my own guidance to its instructions | [Agent settings: System prompt](/agent-settings#system-prompt) |
| Add my policies, FAQs or product documents | [Knowledge](/knowledge) |
| Try the assistant myself before customers do | [Managing your agents: Preview](/managing-agents#testing-an-agent-with-preview) |
| Put the assistant on my website | [Website assistant](/website-assistant) |
| Install the WordPress plugin and keep it updated | [WordPress plugin](/plugin) |
| Read, or listen to, a conversation | [Calls](/conversations) |
| See how many calls it took and how they went | [Dashboard](/dashboard) |
| Have the assistant answer my phone number | [Phone numbers](/phone-numbers) |
| Connect WhatsApp | [WhatsApp](/whatsapp) |
| Register my WhatsApp templates | [WhatsApp templates](/whatsapp-templates) |
| Send a WhatsApp message to one customer | [Sending a WhatsApp message](/whatsapp-send) |
| Import my customer list | [Importing contacts](/importing-contacts) |
| Record a customer's consent, or stop contacting someone | [Contacts](/contacts) |
| Call a group of customers, or send them a WhatsApp message | [Campaigns](/campaigns) |
| Pause, cancel or read the results of a campaign | [The campaign page and its results](/campaign-results) |
| Call customers automatically when something happens | [Automations](/automations) |
| Confirm my store may contact its customers | [Calling consent and calling hours](/calling-hours) |
| Set the hours and days the assistant may call | [Calling consent and calling hours](/calling-hours#the-calling-hours-and-days-card) |
| Know who is never contacted, and why | [How calling rules protect your customers](/outbound-rules) |
| Decide a refund a shopper asked for | [Refund requests](/refunds) |
| Connect my store so the assistant can look up orders | [Connections](/connections) |
| Add a teammate | [Members](/members) |
| Know who can change what | [Who can do what](/roles) |
| See my plan, invoices or cancel | [Plans & billing](/billing) |
| Know my limits and what happens when I reach one | [Plans, limits and what happens at the limit](/limits) |
| Connect my own software | [API keys](/api-keys) and [Webhooks](/webhooks) |
| Look up a word I do not know | [Glossary](/glossary) |
| Find an answer to a common question | [FAQ](/faq) |
## Getting started
What happens on a call, what is managed for you and what you control.
A first-day checklist: from logging in to inviting your team.
Your calls, talk time, success rate and the notices that need you.
The words you meet in the portal, in plain terms.
## Where to get help
Look for your question in the [FAQ](/faq) first. Every page also ends with its own common questions. If you are
still stuck, email us at [info@ruleinside.com](mailto:info@ruleinside.com).
When you write to us about a call, include the agent's name, its **Agent ID** (click it in the agent editor's header
to copy it), the date and time of the call, and what you expected to happen. That helps us find it quickly.
# Knowledge
Source: https://docs.ruleinside.com/knowledge
The documents your agent answers from: adding them, how the agent uses them, and RAG settings for large documents.
Open **Knowledge** to manage what the agent chosen in the **Agent** selector can refer to during a conversation:
web pages, files and text such as your policies, FAQs and product details. Everything on this page applies to that
agent; to work on another agent, choose it in the selector.
## Who can change knowledge
Any member can see this page. To change anything you need **both**:
* to be an **owner** of the account, and
* the active agent to be at **Full control** (see [Edit levels](/agent-settings#edit-levels)). Agents you create
are always at Full control; your main agent often is not.
Otherwise the page is read-only. The header says "This agent is managed, its knowledge is configured by
RuleInside." (agent below Full control) or "Only account owners can change knowledge, ask an owner of your team."
(member), and each document shows its mode as plain text.
## The document list
The page title is **Knowledge Base**, with "What … can reference during a conversation." and your agent's name.
* **Configure RAG** and **Add document** (only when you can change knowledge).
* A search box, **Search Knowledge Base…**, which matches document names.
* Filters: **All**, **File**, **URL**, **Text**.
The search and filters appear once the agent has documents. The list shows:
1. **RuleInside's documents**, marked **Managed**. You cannot change or remove them.
2. **Your documents attached to this agent**, each with a type badge and, when you can change knowledge:
* a mode menu, **Prompt** or **Auto** (see below),
* **Remove**, to take it off this agent only,
* a trash button, to delete it everywhere.
## How the agent uses a document
Each of your documents has a mode on each agent:
* **Prompt**: always given to the agent.
* **Auto**: looked up when it is relevant to the question.
The page footer reads: "Documents in Prompt mode are always included; in Auto mode they're retrieved when relevant
(enable RAG for large documents). Small documents (under \~500 bytes) can't be RAG-indexed."
Switching a document to **Auto** does not prepare it for look-up on its own. Open **Configure RAG**, make sure
**Enable RAG** is on, and click **Save**: that prepares every one of your **Auto** documents on this agent.
## Buttons on a document
| Button | What happens | Can it be undone? |
| - | - | - |
| Mode menu (**Prompt** / **Auto**) | Changes how this agent uses the document. Saves straight away. | Yes, change it back. |
| **Remove** | Takes the document off this agent only. It stays on your other agents and in your documents. Saves straight away. | Yes: attach it again with **Add document**. |
| Trash (**Delete**) | Deletes the document for good and removes it from every agent that uses it. Asks for confirmation first. | **No.** |
Deleting asks "Delete this document?": "'…' will be permanently deleted and removed from every agent that uses it.
To take it off only this agent, use Remove instead. This can't be undone." Click **Delete document**
("Deleting…") to confirm, or **Cancel**.
Deleting a document cannot be undone, and it removes the document from every agent at once. To take it off one
agent only, use **Remove**.
Messages you may see:
* Mode or Remove failed: "Could not update documents." (or the reason), and the page reloads.
* On your main agent below Full control: "This agent is managed, documents can only be attached to your own
agents."
* Delete: "Document deleted." or "Could not delete the document."
## Adding a document
Click **Add document**. A menu opens with two ways to add one.
**Reuse a document you already have.** Search with "Search documents…" among your documents that are not on this
agent (each shows its name and ID), and click one to attach it in **Prompt** mode. "No existing documents. Create
one below." means you have none to reuse; "No matches." means your search found nothing.
**Create a new document** with **URL**, **Files** or **Text**:
| Dialog | Fields | Rules |
| - | - | - |
| **Add URL**: "The agent will read this page as a knowledge source." | **Name (optional)** (placeholder "e.g. Return policy", up to 120 characters); **URL** (placeholder "[https://yourstore.com/faq](https://yourstore.com/faq)") | URL required ("Enter a URL."). "Must be a public https URL, internal/private addresses are blocked." No IP addresses, `localhost` or user names in the address. The name defaults to the URL. |
| **Create text**: "Paste text the agent can use to answer questions." | **Name** (up to 120 characters); **Content** (placeholder "Paste your policy, FAQ, product details…") | Content required ("Enter some text."). If you leave the name empty, the document is called "Text document". |
| **Add file**: "PDF, TXT, DOCX, HTML or EPUB, up to 21MB." | **Name (optional)**; **File** | A file is required ("Choose a file."). Accepted: .pdf, .txt, .docx, .html, .epub. Larger than 21 MB is refused ("File exceeds the 21MB limit."). The name defaults to the file name. |
Click **Add document** ("Adding…") to create it, or **Cancel**. A new document belongs to your account and is
attached to the active agent in **Prompt** mode: "Document added." On failure you see the reason, or "Could not add
the document." ("Could not upload the file." for files).
## Configure RAG
RAG (Retrieval-Augmented Generation) lets the agent pull only the relevant parts of large documents into each
answer, instead of loading whole documents. Click **Configure RAG** to open **RAG configuration**. It is available
only to owners, on an agent at **Full control**.
| Field | What it does | Values | Default |
| - | - | - | - |
| **Enable RAG** | "Retrieval-Augmented Generation (RAG) increases the agent's maximum Knowledge Base size. The agent will have access to relevant pieces of attached Knowledge Base during answer generation." | On or off | The agent's current setting (off if never set) |
| **Embedding model** | How text is matched to the question | **English optimized** or **Multilingual optimized** | English optimized |
| **Character limit** | "Limit the total number of characters retrieved per query." | A whole number, 0 or more | 50000 |
| **Chunk limit** | "Limit the number of document chunks retrieved per query." | A whole number, 1 or more | 20 |
| **Vector distance limit** | "Document chunks with a vector distance higher than this value (i.e., less similar) will not be retrieved." A slider from **More similar** to **Less similar**. | 0 to 2, in steps of 0.01 | 0.6 |
| **Number of candidates** | "Number of candidates evaluated in ANN vector search. Higher number means better results, but higher latency. Minimum recommended value is 100. If disabled, the default value is used." | A switch, then a number of 100 or more | Off; 100 when switched on |
| **Query rewrite prompt override** | "Override the default prompt used for rewriting user queries before retrieving documents. The conversation history will be automatically appended at the end." | A switch, then a text box filled with the default prompt | Off |
While **Enable RAG** is off, only the switch is shown.
Click **Save** ("Saving…"):
* **With RAG on**, your settings are saved, then every one of your documents attached to this agent in **Auto**
mode is prepared for look-up. You see "RAG enabled, indexing … document(s) for retrieval." or "RAG settings
saved." Documents under about 500 bytes, and any that fail, are skipped.
* **With RAG off**, only RAG is switched off; the other values are not saved.
* On failure you see the reason, or "Could not save RAG settings." On your main agent below Full control: "This
agent is managed, RAG can only be configured on your own agents."
## What you might see
| Message | What it means |
| - | - |
| "Knowledge base isn't enabled on your plan. …" | Knowledge is not switched on for your account. Contact RuleInside. |
| "This workspace is not linked to a store yet." | Your workspace is not connected to your store yet. |
| "Couldn't load your knowledge base. Refresh to try again." | The page failed to load. Refresh it. |
| "No documents found" / "This agent has no attached documents yet." | The agent has no documents. Owners also see **Add document**. |
| "No documents match your search." | Nothing matches your search or filter. |
| "You do not have permission to manage knowledge." | You are a member; ask an owner. |
Document controls are greyed out while a change or delete is running.
## Common questions
**Prompt** documents are always given to the agent. **Auto** documents are looked up when relevant; for large
documents this needs RAG switched on and saved.
Open **Configure RAG**, make sure **Enable RAG** is on, and click **Save**. That prepares your Auto documents.
Very small documents (under about 500 bytes) cannot be prepared; keep those in **Prompt** mode.
**Remove** takes the document off this agent only. Delete (the trash button) erases it from every agent, and
cannot be undone.
Either the active agent is below **Full control** (often your main agent) or you are not an owner.
No. Documents marked **Managed** are kept by RuleInside.
PDF, TXT, DOCX, HTML and EPUB, up to 21 MB each.
Yes. Choose the second agent in the selector, click **Add document**, and pick the document from the list at the
top of the menu.
It must be a public `https://` address, with a domain name rather than an IP address, and no user name or
password in it.
There is no edit button. Add the new version as a new document, then delete or remove the old one.
No. The **Knowledge base** card in the agent editor only lists the documents; on an agent you can change it links
here (**Manage knowledge**). Add and manage documents on this Knowledge page.
# Plans, limits and what happens at the limit
Source: https://docs.ruleinside.com/limits
Every limit your account can reach, and what RuleInside does when you reach it.
This page lists the limits that apply to your account and what happens when one is reached. For prices and what
each plan includes, see [ruleinside.com](https://ruleinside.com) or ask our team.
## Minutes
Your plan includes a number of minutes each month, on annual plans too. If you use more, the extra time is
billed as extra usage. Your assistant keeps working: nothing is paused because the included minutes ran out.
You buy a bundle of minutes. Your portal tells you when you have used 80% of it and again at 100%.
When a bundle runs out, the same bundle is bought again automatically with your saved card. This happens at
most twice in one billing period; after the second extra bundle, no further bundle is bought automatically in
that period. A bundle is bought again only when a card is saved on your account and the 80% or 100% notice has
reached you.
Each batch of minutes credited to a pay-as-you-go account is valid for 12 months from the day it is credited.
Minutes you have not used when your subscription ends are not refunded.
## Calls and messages to your customers
These limits protect the people you contact and apply to every account. See
[How calling rules protect your customers](/outbound-rules) for all the checks.
| Limit | Value |
| - | - |
| Calls to one person | At most 5 in any 7 days, counting calls that were not answered |
| WhatsApp marketing messages to one person | At most 4 in any 7 days |
| Calling hours | 8:00 to 21:00 in the person's own time zone, unless you set other hours |
| Attempts per person in one campaign | At most 3, at most 2 in one day, at least 1 hour apart (3 hours unless you choose otherwise, up to 7 days) |
## Daily calls for your account
Each account can place a set number of outbound calls in any 24 hours:
| Plan | Calls in 24 hours |
| - | - |
| Lite | 500 |
| Starter | 1,500 |
| Growth | 5,000 |
| Scale | 15,000 |
| Pay-as-you-go | 1,500 |
When a campaign would go over this number, the people beyond it are not called at that moment. What happens next
depends on where they were:
* **People left out when you launch the campaign** are tried again by that campaign, up to one day of your cap per
launch, starting an hour later or when their calling hours next open, for up to about a day. The launch result
says how many will not be tried again.
* **People being retried, or waiting for their calling hours**, are tried again about an hour later, for up to about
a day.
* Anyone still not reached after that is not called by that campaign.
## Campaigns
| Limit | Value |
| - | - |
| People in one campaign | Up to 5,000. A larger group is cut to the first 5,000. An uploaded list holds at most 1,000 rows |
| Calls at the same time | Lite 3, Starter 5, Growth 10, Scale 20, Pay-as-you-go 5. A higher setting is lowered to your plan's number |
## Your website assistant
Your website assistant can start a limited number of conversations each day, set from your plan's minutes:
| Plan | Conversations a day |
| - | - |
| Lite | 330 |
| Starter | 660 |
| Growth | 1,450 |
| Scale | 2,110 |
The day resets at midnight UTC. Your portal tells you when half of the day's conversations are used and when all
of them are; after that, new conversations start again the next day. Contact us if your store needs more.
## Files and lists
| Limit | Value |
| - | - |
| Contacts imported in one go | 1,000 |
| Contact or campaign file uploaded from the portal | 8 MB |
| One knowledge base file | 21 MB |
## Agents
There is no limit on how many agents you create. Your plan sets how many agents our team builds and manages for
you.
## When a payment fails or your subscription ends
* **A payment fails:** your account keeps working while the payment is retried.
* **It stays unpaid after the retries, or the subscription is paused:** your account is paused. Campaigns cannot
start, WhatsApp messages cannot be sent, and the website assistant stops. Opting people out still works.
* **You cancel:** your plan runs until the end of the period you have paid for. When the subscription ends, minutes
you had not used end with it.
# Managing your agents
Source: https://docs.ruleinside.com/managing-agents
Find, create, test and delete your voice agents, and choose which agent the portal is working on.
Your account can have several voice agents. This page covers the agents list, creating and deleting agents, the
agent selector, the editor's header and testing an agent with **Preview**. Every setting inside the editor is
described in [Agent settings](/agent-settings).
## Who can do what
* **Members** can open the agents list and every agent, and read all of their settings.
* **Owners** can also create, change, test and delete agents. Members see "Only account owners can add or delete
agents, ask an owner of your team." instead of the **New agent** and delete buttons.
If a member somehow starts an action that changes something, it is refused with "Permission denied: this action
requires an account owner (billing) permission." See [Who can do what](/roles).
What an owner can change on a given agent also depends on its **edit level** (**View only**, **Name, voice &
greeting**, **Instructions** or **Full control**). Agents you create are always at **Full control**. See [Edit
levels](/agent-settings#edit-levels).
## Opening your agents
Clicking **Agents** in the sidebar normally opens an agent's editor straight away, not a list. It opens the agent
you last chose in the agent selector, otherwise your main agent, otherwise the first one. If the agent you last
chose no longer exists, it opens your main agent.
You see the full list instead when:
* you choose **Manage agents…** in the agent selector,
* your account has no agents yet, or
* the list could not be loaded, or your workspace is not yet linked to a store.
## The agent selector
The **Agent** selector sits under the page header on the agent editor, **Tools** and **Knowledge** pages. It shows
the agent you are working on and decides which agent those pages show and change.
* The list shows every agent of your account; a tick marks the active one.
* Your main agent, the one RuleInside set up for you, carries a **Managed** badge (in the list it reads
"managed"). The badge only marks your main agent. What you can change on it is decided by its edit level.
* On the agent editor, choosing another agent opens that agent's editor. On **Tools** and **Knowledge**, it keeps you
on the same page and reloads it for that agent.
* **Manage agents…** at the bottom opens the full agents list.
* Your choice is remembered in this browser for 30 days, so your next visit opens the same agent.
* The selector is hidden when your workspace is not linked to a store, when there are no agents, or when the list
could not be loaded.
## The agents list
"Your voice agents. Open one to edit it, or create a new one."
Each row shows:
* the agent's name (click the row to open its editor),
* a **Managed** badge on your main agent,
* the agent's edit level under its name,
* a trash button to delete it (owners only, and never on your main agent).
| What you see | What it means |
| - | - |
| "No agents yet. Create your first one." | Your account has no agents. |
| "This workspace is not linked to a store yet." | Your portal workspace is not connected to your store yet. **New agent** is greyed out. Contact RuleInside. |
| "Couldn't load your agents. Refresh to try again." | The list failed to load. Refresh the page. |
## Creating an agent
Click **New agent** (owners only). The dialog reads "Start from scratch or clone an existing agent. You can fully
control a new agent's prompt, voice and settings."
| Field | What it does | Values and limits | Default |
| - | - | - | - |
| **Blank** ("Author from scratch") / **Clone** ("Copy an existing agent") | How the new agent starts | One of the two. **Clone** is greyed out when you have no agents. | Blank |
| **Name** | The agent's name (placeholder "e.g. After-hours assistant") | Required, 1 to 100 characters. An empty name shows "Give your agent a name." | |
| **Clone from** (Clone only) | The agent to copy. "The new agent inherits the source's prompt, voice, knowledge base and tools." | Any of your agents (your main agent shows "(managed)") | |
| **Language** (Blank only) | The starting language | A short language code such as `en` or `en-gb` | `en` |
| **First message** (Blank only) | The starting greeting (placeholder "Hi! How can I help you today?") | Optional, up to 1,000 characters | Empty |
| **Instructions (optional)** (Blank only) | What the agent should do and how it should behave | Optional, up to 8,000 characters | Empty |
Then click **Create agent** ("Creating…" while it works), or **Cancel** to close without creating anything.
* On success you see "Agent created." and the new agent's editor opens.
* On failure you see the reason, or "Could not create the agent."
* A new agent is never your main agent, and it is always at **Full control**.
What the new agent starts with:
* **Blank:** the language model RuleInside sets for new agents (you cannot change it), and the instructions you
typed (or "You are a helpful voice assistant." if you left them empty). The system prompt shows a first line `/no_think` above your instructions.
The first message and language are set only if you filled them in. You choose the voice in the editor afterwards.
* **Clone:** a copy of the source agent's prompt, voice, knowledge and tools. If **Transfer to a person** is not
switched on for your account, transfer numbers are not copied.
Type the language as a code, such as `en`, `it` or `en-gb`. A name like "English" is refused, and the dialog then
shows only "Could not create the agent."
There is no plan limit on how many agents you create. If RuleInside has set a limit for your account, creating is
refused with "An agent limit of N is set on this account. Contact support to raise it."
If you see "Could not secure the new agent. Nothing was created, please try again.", nothing was created; try
again.
## Deleting an agent
Click the trash button on an agent's row (owners only; your main agent has none). The confirmation reads "Delete
this agent?" and "'…' will be permanently removed. Calls already logged are kept. This cannot be undone." Click
**Delete agent** ("Deleting…" while it works) to confirm, or **Cancel**.
* On success you see "Agent deleted." On failure you see the reason, or "Could not delete the agent."
* Your main agent cannot be deleted: "Your primary agent is managed and cannot be deleted."
Deleting an agent cannot be undone.
Before you delete an agent, check whether a website assistant or a phone number uses it. See [Website
assistant](/website-assistant) and [Phone numbers](/phone-numbers).
## The editor header
At the top of an agent's editor:
* **The agent's name** (or "Your agent" if it has none), with "Voice assistant for your store" under it.
* **Agent ID**: the agent's full ID. Click it to copy it (a tick confirms; the tooltip reads "Copy agent ID"). This
is the ID you paste into the website assistant embed code.
* **Preview** (owners only): test the agent, see below.
* **Your plan**, for example "Starter plan". For information only.
* **The edit level badge**: what you can change on this agent.
If you open an agent that does not belong to your account (for example from an old bookmark), you see a "not found"
page.
## Testing an agent with Preview
Click **Preview** (owners only) to talk to the agent in your browser. The panel opens with the agent's name in its
title ("Loading preview…" while it loads).
Click **Text** or **Voice**. For voice, your browser asks for permission to use the microphone: "Voice uses
your microphone, your browser will ask for permission."
Click **Start text chat** or **Start voice call** ("Starting…"). The status changes from "Not connected" to
"Connecting…" and "Connected". In voice, it shows "Agent speaking…" or "Listening…".
In text, type in "Type a message…" and send ("Say hello to start." while the chat is empty). In voice, just
talk: the bottom of the panel shows "Agent is speaking…" or "Listening, just talk".
Click **End**. Closing the panel also ends the session.
Good to know:
* Preview talks to the real agent with its **saved** settings. Click **Save changes** first if you want to test a
change.
* Preview does not identify you as a customer logged in to your store, so store tools answer as they would for a
guest. A guest gets an order's details only by giving its order number with the billing email and postcode on it.
* If something goes wrong you see "Could not start the preview.", "Could not connect.", "Preview error." or
"Conversation ended:" followed by the reason ("connection error" when no reason is given).
## Common questions
There is no plan limit. A limit applies only if RuleInside has set one for your account, and none is normally
set.
Your main agent, marked **Managed**, is protected. Only agents you created can be deleted.
No. Calls already logged are kept. The agent itself cannot be brought back.
That is how the page works: it opens the agent you were last working on. To see the list, choose **Manage
agents…** in the agent selector.
Not in the **New agent** dialog. Create the agent, then choose its voice in the editor. See
[Voice](/agent-settings#voice).
It marks your main agent, the one RuleInside set up for you. It does not by itself lock anything: what you can
change depends on the agent's edit level.
Only account owners see them. Ask an owner of your team.
In the agent's editor header, click **Agent ID** to copy it.
Those pages follow the agent selector. Choose the agent you want in the **Agent** selector at the top of the
page.
# Members
Source: https://docs.ruleinside.com/members
Invite your team, manage pending invitations, change roles, remove people and transfer ownership.
The **Members** page shows who belongs to your organization and which invitations are still waiting. Open it from
the left sidebar, under the **Settings** heading. Everyone in the organization can open it. What each person can do
on it depends on their role: see [Who can do what](/roles).
## Organization Members
The first card, **Organization Members**, lists everyone with access.
**Search members** filters the list as you type. It matches a person's name and their role (type `owner` to see
the owners), and ignores upper and lower case. It does not match the part of an email after the `@`.
| Column | What it shows |
| - | - |
| **Name** | The person's picture and name. If they have not set a name, the part of their email before the `@`. Your own row carries a **You** badge. |
| **Email** | The email they sign in with. |
| **Role** | **Owner** or **Member**. The primary owner also carries an amber **Primary Owner** badge. |
| **Joined at** | The day they joined, in your browser's date format. |
Owners are listed before members. When there is nobody to show, the table reads **No data available**.
### The actions menu
The **...** button at the end of a row opens the actions you are allowed to take on that person. It never appears
on your own row or on the primary owner's row, and it does not appear at all if you may take none of the actions.
| Menu item | Who sees it |
| - | - |
| **Update Role** | The primary owner, on anyone. Any other owner, on members only. |
| **Transfer Ownership** (red) | Only the primary owner. |
| **Remove from Account** (red) | The primary owner, on anyone. Any other owner, on members only. |
## Inviting people
Everyone in the organization sees the **Invite Members** button, at the top right of the members card.
Click **Invite Members**. The dialog **Invite Members to your Organization** opens. It may show "Loading.
Please wait..." for a moment first.
Type the person's email and choose their role. Click **Add another one** for each extra person, up to 5 in one
go. Remove a row with its **X** button (**Remove invite**).
Click **Send Invites**. You see "Members invited successfully", and each person receives an email with a link
to join.
| Field | What it does | Allowed values | Default |
| - | - | - | - |
| Email | The address the invitation goes to. It is saved in lower case, without spaces. | Any valid email. Required. | Empty |
| Role | The role the person gets when they accept. | **Member** or **Owner**. You can only offer roles up to your own, so a member can invite only members. | **Member** |
Good to know:
* The same address twice in one form is refused with "Duplicate emails are not allowed".
* If **any** address in the form already belongs to a member of your organization, the whole form is refused and
nobody is invited. You see "Sorry, members could not be invited. Please try again." Take that address out and
send again.
* An address that already has a pending invitation to your organization cannot be invited again. Remove the old
invitation first, or renew it if it has expired.
* An invitation is valid for **7 days**.
* The person must accept with the same email you invited (upper and lower case do not matter). If they sign in
with a different address, the link tells them the invitation was not found. See
[Accepting an invitation](/your-account#accepting-an-invitation).
## Pending Invites
The second card, **Pending Invites**, lists invitations that have not been accepted yet. Everyone can see it.
**Search Invitations** matches the part of the email before the `@`, and the role.
| Column | What it shows |
| - | - |
| **Email** | Who was invited. |
| **Role** | The role they will get. |
| **Invited at** | When the invitation was sent. |
| **Expires at** | When the link stops working. |
| **Status** | **Active** (green): the link still works. **Expired** (amber): the 7 days have passed. |
An accepted invitation disappears from the list. An expired one stays until someone renews or removes it.
### Managing an invitation
Owners see a **...** menu on each invitation. Members do not.
* **Update Invitation** changes the role the person will get. The dialog is titled **Update Member's Role**, and
you save with **Update Role**.
* **Renew Invitation** appears only on an **Expired** invitation. It makes the link the person already has work
again for 7 days from now. **It does not send a new email**: tell the person to use the email they already
received.
* **Remove Invitation** opens **Delete Invitation**. Press **Delete Invitation** to confirm. The emailed link stops
working at once. This cannot be undone: to let the person in later, invite them again.
Only the primary owner can change or renew an invitation for the **Owner** role, or turn a member invitation
into an owner one. Any other owner manages member invitations only.
## Changing someone's role
On the person's row, open the **...** menu and choose **Update Role**. The dialog **Update Member's Role**
opens.
In **Role**, choose **Owner** or **Member**. Their current role is greyed out. Picking the same role shows
"Role must be different from the current one".
Click **Update Role**, or **Cancel** to close without saving.
The change takes effect at once. A member you promote to owner can immediately manage billing and every other
owner-only control. Nobody can change their own role, and nobody can give a role higher than their own.
## Removing someone
On the person's row, open the **...** menu and choose **Remove from Account**. The dialog **You are removing this
user** asks you to confirm. Click **Remove User from Organization**.
* The person loses access to this organization straight away.
* Their login is **not** deleted. They can still sign in, and they keep any other organizations they belong to.
* To bring them back, invite them again.
* The primary owner cannot be removed.
## Transferring ownership
Only the primary owner can do this. Use it to hand the organization to someone else, or before you leave or delete
your own login.
On their row, open the **...** menu and choose **Transfer Ownership**.
Click **Send Verification Code**. We email a 6-digit code to your own address. Type it in and click
**Verify Code**. If the email does not arrive, click **Request New Code**. You get at most 5 tries per code.
A red box names the person you are transferring to. Click **Transfer Ownership** to finish, or **Cancel** to
stop.
What changes:
* The person you chose becomes the primary owner, and gets the **Owner** role if they did not have it.
* You keep the **Owner** role and stay in the organization. You can now leave it if you wish.
* Only the new primary owner can transfer ownership back.
## Common questions
Ask them to check their junk folder first. If the email still has not arrived, remove the invitation and invite
them again. There is no resend button, and **Renew Invitation** does not send an email.
The invitation is more than 7 days old, was removed, or was already used. It also happens when the person signs
in with a different email from the one you invited. If the invitation shows **Expired**, an owner can click
**Renew Invitation** and the same link works again. Otherwise, invite the address they actually sign in with.
Yes, but only with the **Member** role. Members cannot change or remove invitations afterwards.
As primary owner, use **Transfer Ownership** (you will need the emailed code). Afterwards you can leave from
[Organization settings](/organization-settings#leaving-the-organization).
No limit, and no extra cost. No plan is billed per person.
Invite them again. Their login still exists, so they can accept straight away.
You may not act on that person. The menu never shows on your own row or on the primary owner's row, and an
owner who is not the primary owner sees it only on members.
# Organization settings
Source: https://docs.ruleinside.com/organization-settings
Your organization's name and logo, how renaming changes your web address, and leaving an organization.
Open **Settings** from the left sidebar, under the **Settings** heading. The tabs along the top read **Settings**,
**Profile**, **Plugin**, **Calling consent** and **Developer**. This page covers the first tab, **Settings**, which
holds your organization's logo and name and the option to leave. Your personal details are on the **Profile** tab:
see [Your account](/your-account).
## Who can change these settings
* Everyone in the organization can open this tab.
* **Only the primary owner can save a new name or logo.** Anyone else can type in the form, but saving shows
"Could not update Organization. Please try again." See [Who can do what](/roles).
* Leaving is open to everyone **except** the primary owner.
## Organization Logo
The **Organization Logo** card sets the picture that identifies your organization. It appears, for example, on
the page where an invited colleague accepts their invitation.
* Click the uploader and choose an image file. The uploader's text reads **Upload a Profile Picture**: it is the
same control as on your personal profile, but here it sets the organization's logo.
* Choosing a new image replaces the old one. Clearing it removes the logo.
* You see "Updating Organization..." and then "Organization successfully updated".
## Organization Name
The **Organization Name** card holds one field, **Organization Name**, and the **Update Organization** button.
| Rule | What you see if it is broken |
| - | - |
| 2 to 50 characters. Required. | A message about the length. |
| None of these characters: `! @ # $ % ^ & * ( ) + = [ ] { } ; ' : " \ \| , . < > / ?` | "This name cannot contain special characters. Please choose a different one." |
| Not `settings` or `billing`, in any mix of upper and lower case. | "This name is reserved. Please choose a different one." |
Accented letters such as é or ü are fine.
### Names in other scripts
If the name contains Cyrillic, Hebrew, Arabic, Thai, Korean, Japanese or Chinese characters, a second field
appears: **Organization URL**. It sets the short name used in your web address, and it is required in that case.
* Only lowercase English letters (a to z), numbers (0 to 9) and single hyphens, 2 to 50 characters. Example:
`my-organization-name`.
* It cannot be `settings` or `billing`.
* It must not already be taken. Otherwise you see "This URL is already taken. Please choose a different one."
### Renaming changes your web address
Your organization's web address is built from its name, for example `/home/my-store/...`. **When you rename the
organization, the web address changes too.** After you save, the portal moves you to the new address. Bookmarks
and links to the old address stop working, so update them and tell your team.
If the new address is already used by another organization, a number is added to the end of it.
## Danger Zone
The **Danger Zone** card holds actions that cannot be undone. What you see depends on your role.
### Leaving the organization
Owners and members who are not the primary owner see **Leave Organization**.
The dialog **Leaving Organization** opens.
In the field **Please type LEAVE to confirm leaving the organization.**, type `LEAVE` in capital letters.
Click **Leave Organization**, or **Cancel** to stay.
You lose access at once. The portal takes you to another organization you belong to or, if you have none, to the
page for creating a new organization. **This cannot be undone**: to come back, an owner must invite you again.
### If you are the primary owner
The primary owner has no **Leave Organization** option. To leave, first hand the organization to someone else with
**Transfer Ownership** on the [Members](/members#transferring-ownership) page. You then keep the owner role and can
leave like anyone else.
## Common questions
Only the primary owner can. Check the [Members](/members) page for the person with the **Primary Owner** badge
and ask them, or ask them to transfer ownership to you.
That is expected: the web address follows the name. Open the portal from the start page and bookmark the new
address.
**Settings**, **Danger Zone**, **Leave Organization**, then type `LEAVE`. If you are the primary owner, transfer
ownership first.
Cancel your plan on [Billing](/billing#cancelling-your-plan). Leaving the organization or deleting your own
login does not stop billing. Contact us if you need anything else closed.
Yes. The logo and the name are saved separately. Only the name changes your web address.
# How calling rules protect your customers
Source: https://docs.ruleinside.com/outbound-rules
The checks every outbound call and WhatsApp message passes before anyone is contacted.
Every time your assistant is about to call someone, or send them a WhatsApp message, RuleInside checks that person
first. The checks run for each person, every time, and a person who fails any one of them is not contacted. You do
not have to switch them on: they apply to every account.
## The checks
Before any of the checks on a person, RuleInside checks your whole account:
* **Your store has confirmed it may contact its customers.** An owner confirms, under
[Settings, Calling consent](/calling-hours), that the store has a lawful basis to have an automated AI voice agent
contact the people it asks RuleInside to contact, in the country it trades in. Until this is confirmed, **no calls
and no WhatsApp messages** are sent.
Then, for each person, in this order:
Someone who opted out on a channel is never contacted again on that channel. People opt out by telling your
assistant during a call, by replying STOP on WhatsApp, through your own system, or when someone in your team uses
**Opt out** on the [Contacts](/contacts) page (which stops both calls and WhatsApp). Only the customer can
reverse an opt-out, and only on WhatsApp, by replying START (see [WhatsApp](/whatsapp)).
On WhatsApp there is also a marketing-only stop: someone who told WhatsApp to stop marketing messages from your
business is sent no more **Marketing** templates from your store, until they turn your marketing messages back
on in WhatsApp. See [Why a send is suppressed](/whatsapp-send#why-a-send-is-suppressed).
Numbers on your do-not-call list are never contacted on that channel. Every opt-out is added to the list
automatically, so it stays in force even if the person is imported again later. Calls and WhatsApp are listed
separately: a STOP on WhatsApp stops both, while a "do not call me" on the phone does not stop WhatsApp.
A person is called only when your contact list records a lawful basis to call them, and is sent a WhatsApp message
only when it records a lawful basis to message them on WhatsApp. One channel never implies the other. A newly
imported contact has no consent until you record it (see [Importing contacts](/importing-contacts)).
Calls are placed between 8:00 and 21:00 in the person's own time zone, on any day, unless you set other calling
hours or days for your account or for one campaign. Hours outside 8:00 to 21:00 must be confirmed by an owner.
The time zone comes from the contact's details or their phone number; when neither gives one, the person is not
called unless you have set a **Timezone for customers we cannot place**. See
[Calling hours](/calling-hours#the-calling-hours-and-days-card). WhatsApp **Marketing** templates and a campaign's
after-call WhatsApp message also wait for the person's calling hours; other WhatsApp messages do not. A Marketing
follow-up in a chat the shopper started goes at once only within 24 hours of that chat.
RuleInside limits how often one person can be contacted, and each account has a maximum number of calls per day.
The numbers are on [Plans, limits and what happens at the limit](/limits).
## When a check cannot be completed
If a check cannot be completed at that moment, for example because a list could not be read, the person is **not**
contacted. RuleInside never contacts someone because a check failed to answer. In the same way, if your saved calling
hours or days could not be read, nobody is called.
When calls and messages are held back only because your store has not confirmed calling consent yet, they are not
cancelled at once: they are tried again a limited number of times. Confirm before you launch a campaign so that
nothing is held back.
## Campaigns with their own hours
A campaign can use its own calling hours, its own calling days, or both, instead of the account's. They can be changed
only while the campaign is a draft. Each person in the campaign is still checked against every other rule above. See
[Campaigns](/campaigns).
These rules are the minimum RuleInside applies. You remain responsible for the consent you record and for the
rules of the countries you call. RuleInside does not check that the lawful basis you record is true.
# Phone numbers
Source: https://docs.ruleinside.com/phone-numbers
Import a Twilio number, choose which agent answers it and which one calls out, release a number, and connect WhatsApp.
Open **Phone numbers** to bring your Twilio phone numbers into RuleInside, choose which agent answers calls to each
number and which agent places outbound calls from it, and manage your store's WhatsApp number. A
[campaign](/campaigns) needs a number here that can place outbound calls.
## Who can change what
* **Every member** can see the page.
* **Only an owner** can import, assign or release a number, or change WhatsApp settings. A member sees **"Only
account owners can manage this - ask an owner of your team."** in place of the **Import number** button, and
**Assign** and **Release** are greyed out. If a member's request reaches the server anyway, it is refused with
"Permission denied: this action requires an account owner (billing) permission." See [Who can do what](/roles).
The **Agent** bar at the top of the page changes nothing here: phone numbers belong to your store, not to one agent.
## Your numbers
The **Phone numbers** card has the subtitle **"Import a Twilio number, then assign the agent that answers inbound
calls and the one that places outbound calls."**
| Column | What it shows |
| - | - |
| **Label** | The name you gave the number. A dash if it has none. |
| **Number** | The phone number. |
| **Provider** | Where the number comes from, for example "Twilio". |
| **Inbound agent** | The agent that answers calls to this number. A dash means **no agent answers** calls to it. |
| **Outbound agent** | The agent that places outbound calls from this number, for example in a campaign. A dash if none. |
| (buttons) | **Assign** opens the agent choice for the number. **Release** (red) removes the number. |
| Message | What it means |
| - | - |
| **"No phone numbers yet. Import a Twilio number to get started."** | You have no numbers yet. |
| **"This workspace isn't linked to a store yet."** | Your account is not connected to a store, so **Import number** is greyed out. |
| **"We couldn't load your phone numbers right now. Please refresh - if it keeps happening, contact support. (Your numbers are safe; nothing was changed.)"** (red) | The list could not be loaded. Your numbers are still there: refresh rather than importing them again. |
## Import a number
Owners only. You need the number's **Account SID** and **Auth Token** from your Twilio console.
Press **Import number**. The form **Import a Twilio number** opens, with the note "Your Twilio SID and Auth Token
are sent once to provision the number and are never stored by us. Find them in your Twilio console."
| Field | Required | What to enter |
| - | - | - |
| **Phone number (E.164)** | Yes | The number in international form: a plus sign, the country code and the number, for example `+15551234567`. Spaces are accepted and removed. It must be a valid phone number, or the import is refused before anything is created. |
| **Label** | **Yes** | A name for the number, for example "Support line". The form does not mark it as required, but **Import** stays greyed out until it is filled. |
| **Twilio Account SID** | Yes | From your Twilio console. Hidden as you type. |
| **Twilio Auth Token** | Yes | From your Twilio console. Hidden as you type. |
| **Inbound agent (optional)** | No | The agent that answers calls to the number. Default **Unassigned** (no agent answers). You can choose it later with **Assign**. |
Spaces at the start and end of every field are removed.
Press **Import**. It reads **Importing…** while working. When it finishes you see **"Phone number imported."** and
the number appears in the list. Press **Cancel** to close the form without importing.
Your Twilio Account SID and Auth Token are sent once to set the number up. RuleInside never stores them or writes them
to a log.
If the import fails, nothing is left half-done and you see one of these messages:
| Message | What to do |
| - | - |
| A message that the number is not a valid phone number | Check the number and type it with a plus sign and the country code. |
| "That phone number is already linked to another client." | The number is already in use on another RuleInside account. Contact support. |
| "Could not verify phone-number ownership right now. Please retry." | Try again in a moment. |
| "Could not import the phone number." | The import did not complete. Check the number, the Account SID and the Auth Token, then try again. If it keeps happening, contact support. |
## Assign agents
Owners only. Press **Assign** on a number's row.
The form **Assign agents** opens with the number and the text "choose which agent answers inbound calls and which
places outbound calls."
| Field | What it does |
| - | - |
| **Label** | The number's name, filled in with the current one. Empty it to remove the label. |
| **Inbound agent** | **Unassigned** or one of your agents. The change applies **immediately** when you save: calls to the number are answered by the new agent, or by none if you choose **Unassigned**. Greyed out, with **"This number does not support inbound calls."**, if the number cannot receive calls. |
| **Outbound agent** | **Unassigned** or one of your agents. This is the agent that places outbound calls from the number. Greyed out, with **"This number does not support outbound calls."**, if the number cannot place calls. |
Press **Save** to apply your changes (greyed out until you change something; it reads **Saving…** while working).
On success you see **"Phone number updated."**; otherwise **"Could not update the phone number."** or the reason. Only
your own agents can be chosen: any other is refused with "Cannot assign the phone number to an agent outside this
client." Press **Cancel** to close without saving.
## Release a number
Owners only. Press **Release** (red) on a number's row. A confirmation asks **"Release this phone number?"** and
explains that the number "will be removed from your workspace and will stop routing calls. This cannot be undone."
* Press **Release** to confirm. On success you see **"Phone number released."**
* Press **Cancel** to keep the number.
**Releasing cannot be undone.** The number stops answering and placing calls, and its agents and label are removed.
To use it again, you must import it again with your Twilio Account SID and Auth Token.
If the release fails, you see **"Could not release the phone number."** (or the reason) and the number stays in your
list, so you can press **Release** again.
## WhatsApp
The **WhatsApp** card, below your numbers, shows your store's WhatsApp number. Its description reads "WhatsApp
Business accounts are connected during onboarding. Here you can reassign the agent and turn messaging on or off." An
owner can also connect a number directly from this card, as described below. For templates, sending and the send log,
see [WhatsApp](/whatsapp).
### What the card can show
| What you see | What it means |
| - | - |
| **"No WhatsApp accounts are connected to your agents yet."** | No number is connected. Owners see the **Connect WhatsApp** block. |
| **"This workspace isn't linked to a store yet."** | Your account is not connected to a store. |
| **"We couldn't load your WhatsApp accounts right now. Please refresh - if it keeps happening, contact support."** (red) | The card could not be loaded. Refresh the page. |
| Red text beginning "Your WhatsApp connection is recorded but incomplete, so we can't use it to send or receive anything yet." | A connection exists but cannot be used, and nothing can be sent or received. An owner can reconnect the number with **Connect WhatsApp**, or contact support. A member sees "A workspace owner can reconnect the number, or contact support and we'll sort it out for you." |
### A number connected through Meta
The card shows the number, the badge **Connected through Meta** and the note "This number is connected directly to
Meta. Its settings are managed by us, so there is nothing to change here. To disconnect, use Settings > Account >
Business Platform in your WhatsApp Business app."
Underneath, it shows what WhatsApp reports about sending from the number:
| Status | What you see |
| - | - |
| Can send (green) | **"WhatsApp lets this number send messages."** |
| Can send with a limit (amber) | **"This number can send messages, with a limit:"** followed by Meta's notes, for example that your display name is not yet approved. |
| Blocked (red) | **"WhatsApp is blocking messages from this number. Until it is fixed, the messages you send wait here and are not sent:"** followed by Meta's reasons and how to fix them. |
| Not known (grey) | **"We couldn't check this number's status with WhatsApp right now."** |
While WhatsApp blocks the number, your messages are held and go out once the problem is fixed.
### A number you can edit
For other connected numbers, the card shows the number, the business name, and a red **Token expired** badge when the
connection needs renewing. An owner can change:
| Setting | What it does |
| - | - |
| **Agent** | Which of your agents handles WhatsApp. You must choose an agent: there is no "Unassigned" option. |
| **Messaging enabled** | Switch WhatsApp messaging on or off. |
| **Voice-note replies** | Switch on to have the agent reply with voice notes. |
Press **Save** (greyed out until you change something; **Saving…** while working). You see **"WhatsApp account
updated."** or **"Could not update the WhatsApp account."**
### Connect WhatsApp
Owners only. The **Connect your WhatsApp number** block appears when no number is connected yet, or when the existing
connection is incomplete. If it is not available for your account, the card reads "WhatsApp self-connection isn't
switched on for this environment yet. Contact support and we'll connect your number for you."
The block explains: "Sign in with the Meta account that owns your WhatsApp Business number. You'll pick the number
yourself, and we never see your Meta password." Read the four points shown before the button:
* **"You keep using the WhatsApp Business app on your phone, and it has to be opened at least once every 13 days. If
you uninstall it, the connection drops."**
* **"Some things change on your phone for good: disappearing messages and view-once are turned off in one-to-one
chats, live location sharing stops, broadcast lists become read-only, group chats are not carried over, and your
profile photo can no longer be changed."**
* **"Keep the app open for a few minutes after you connect while your existing conversations sync."**
* **"After connecting, add a payment method to your WhatsApp Business Account in Meta. Meta bills you directly for the
messages you start, and blocks them until a payment method is added."**
A Meta window opens and the button reads **Connecting…**. Allow pop-ups for the portal if your browser blocks the
window. Do not reload the portal page while the Meta window is open: the attempt would be lost and you would need
to start again.
Sign in with the Meta account that owns your WhatsApp Business number and choose the number. When you finish, the
Meta window shows "Finishing up. This window will close on its own." and closes.
If your Meta login manages more than one WhatsApp Business account, the portal asks **"Which WhatsApp number is
this store’s?"**. Pick the number your customers message this store on (an account without a name shows as
**Unnamed WhatsApp account**) and press **Connect this number**, or **Cancel**. Nothing is connected until you
confirm.
When you see **"WhatsApp connected."**, add a payment method to your WhatsApp Business account in Meta.
While the Meta window is open, **Start over** lets you abandon the attempt and begin again.
| Message | What it means |
| - | - |
| **"WhatsApp connected."** | Your number is connected. |
| **"You closed the Meta window before finishing. Nothing was changed."** | Press **Connect WhatsApp** to try again. |
| **"Your browser blocked the Meta window. Allow pop-ups for this site, then press Connect WhatsApp again."** | Allow pop-ups for the portal and try again. |
| **"Meta didn't finish the sign-in, so nothing was connected. You can try again."** | Press **Connect WhatsApp** again. |
| **"You had started over, so the connection you just finished in the Meta window was not saved. Press Connect WhatsApp to do it again."** | You pressed **Start over** before the Meta window finished. Connect again. |
| **"Could not complete the WhatsApp connection."** | The connection did not complete. Try again, or contact support. |
| "This page was opened on its own, so there is nothing to finish here. Go back to Phone numbers and press Connect WhatsApp." | You opened the Meta return page directly. Go back to **Phone numbers** and press **Connect WhatsApp**. |
## Common questions
Only owners can import numbers. If you are an owner and the button is still greyed out, your account is not yet
connected to a store.
The **Label** field is required even though the form does not mark it: **Import** stays greyed out until it is
filled. Enter any name that helps you recognise the number, for example "Support line".
No. Your Account SID and Auth Token are sent once to set the number up, and RuleInside never stores them or writes
them to a log.
It must be a real, valid phone number. Type it with a plus sign and the country code, for example `+15551234567`.
Check the number's **Inbound agent**. A dash means no agent answers it: press **Assign** and choose one.
The number cannot place outbound calls; the form shows "This number does not support outbound calls."
None of your numbers can place outbound calls. Import a number that can, or contact RuleInside.
No. Import the number again with your Twilio Account SID and Auth Token.
It is connected through Meta, and its settings are managed by us. To disconnect it, use Settings > Account >
Business Platform in your WhatsApp Business app.
Follow the reason shown on the card; a missing payment method in Meta is a common one. Your messages wait in
RuleInside and go out once it is fixed.
Both are possible: our team can connect your number during onboarding, and an owner can also connect it from this
card with **Connect WhatsApp**.
# WordPress plugin
Source: https://docs.ruleinside.com/plugin
Download the RuleInside plugin for your store, and manage the licence key that keeps it up to date.
Open **Settings** and choose the **Plugin** tab. The **WordPress plugin** card has two parts: **Download**, to get
the plugin for your store, and **Licence key**, the key the plugin uses to receive updates automatically.
## Who can do what
| Action | Who |
| - | - |
| See this page | Everyone on your team |
| **Download plugin** | Everyone on your team, so whoever installs WordPress for you can get it |
| **Create licence key** and **Revoke** | The owner only. Members see "Only an account owner can create or revoke a licence key." |
If your workspace is not linked to a store yet, the page tells you there is nothing to download. Finish your setup
first, or contact us.
## Download the plugin
The **Download** part shows the latest version (for example `v1.5.0`), the date it was **Released**, and the
WordPress and PHP versions it needs (**Needs WordPress** and **Needs PHP**). Open **What changed in this version** to read the
release notes.
Click **Download plugin**. The button reads "Preparing…" for a moment, then your browser downloads the plugin as a
zip file. Each click prepares a fresh download that is valid for 5 minutes, so the button works even if the page
has been open for a long time. Then install the zip in WordPress as you would any plugin.
If the download fails, you see "Could not prepare the download. Please try again." or a more specific message.
Instead of the download, the card may show:
| Title on the card | What it means | What to do |
| - | - | - |
| "We could not check for the latest version" | We could not look up the latest version. The problem is on our side, not with your store. | Try again in a moment, and contact us if it keeps happening. Do not install an older copy instead. |
| "Plugin downloads are paused for this account" | Your RuleInside account is not active, so new downloads are unavailable. A plugin you already installed keeps working. | Contact us. Trying again will not help. |
| "No release is published yet" | There is no plugin version to download at the moment. | Contact us. |
## Customer details from your store
When a customer's name, email, phone or country changes in WordPress, the plugin sends the new details to
RuleInside, so your assistant uses the current ones. Nothing is sent unless the **Widget ID** and **Identity Token
Secret** are both filled in on the plugin's settings page (WordPress admin, menu **RuleInside**).
From version 1.5.1, only customers are sent. A customer is a user whose roles are only WooCommerce's `customer`
and/or WordPress's `subscriber`, and who has none of these staff permissions: writing content, managing WooCommerce,
or FluentCart staff access. Staff accounts, custom staff roles and customers who were also given a staff role are not
sent.
If your shoppers have a renamed or plugin-added role instead of, or as well as, `customer` or `subscriber` (some
forum plugins, such as bbPress, add one to every logged-in user), their profile changes are not sent, and the
assistant keeps the name it already has. Tell us if that applies to your store.
## Licence keys
A licence key lets the plugin installed on your store receive updates automatically. It only unlocks plugin
downloads: it cannot read your data or place calls.
### Create a licence key
Owner only. The button reads "Creating…" while the key is made. The key always has exactly one permission,
plugin downloads; there is nothing to choose.
The **Copy your licence key now** dialog shows the full key, which starts with `ri_sk_live_` (for example
`ri_sk_live_xxxxxx`). Click **Copy** (it changes to **Copied**). If your browser blocks copying, select the key
and copy it by hand.
The key is never shown again, to you or to us. We store only a scrambled fingerprint of it, so nobody can read
it back.
In your WordPress admin, open the **RuleInside** menu and paste the key into the **License Key** field. The
plugin then checks for updates about every 12 hours.
Copy the key before you click **Done**. If you lose it, there is no way to see it again: create a new key, paste it
into the plugin, then revoke the old one.
### The list of licence keys
Each licence key is listed with:
* the start of the key followed by "…" (the full key is never shown again);
* **Created** and the date;
* **Last used** with the date and time, or **Never used**. **Last used** moves each time your store's plugin
checks for updates, so it tells you the key is working;
* **Revoke** (owner only).
With no keys you see "No licence key yet. Create one, then paste it into the plugin's settings screen."
If you see "We could not load your licence keys", the list could not be loaded at that moment. Your existing keys
are unaffected and still work. Try again shortly, and do not create a new key to work around it, or you may end up
with duplicates.
### An API key that already works as a licence
If you see "One of your API keys already works as a licence", one of your [API keys](/api-keys) includes plugin
downloads alongside other permissions, so the plugin accepts it and you do not need another key. If you prefer to
keep the licence separate from your other permissions, create a licence key here, put it in the plugin, and then
revoke the broader key on the **Developer** tab.
### Revoke a licence key
Click **Revoke** on the key and confirm with **Revoke key**. You see "Licence key revoked.", or "That licence key
was already revoked." if it was already gone.
What a revoked key does:
* any store still using it **stops receiving plugin updates** straight away;
* the plugin itself, the website assistant on your store and your calls **keep working**;
* a download your store had already been offered can still be installed for up to 24 hours.
Revoking cannot be undone. To keep your store updating, create a new key and paste it into the plugin.
Licence keys also appear on the **Developer** tab among your [API keys](/api-keys), with a `plugin:download` badge.
Revoking one there has the same effect as revoking it here.
## Facts about licence keys
* A licence key works only while your account is active.
* When the licence is wrong, missing or revoked, WordPress simply shows that no update is available. Your store
shows no error, so check **Last used** on this page instead.
## Common questions
**Settings**, **Plugin** tab, **Download plugin**. Anyone on your team can download it.
The licence key gives your store automatic plugin updates. Paste it into the **License Key** field in the
plugin's settings.
Paste your widget's embed id into the plugin's **Widget ID** field (WordPress admin, menu **RuleInside**). You
find the embed id on the [website widgets](/website-assistant) list and in the widget editor.
It cannot be shown again. Create a new key, paste it into the plugin, then revoke the old one.
Check the key's **Last used** date on this page. **Never used** means the key was never pasted into the plugin,
or was pasted wrongly. Also check that the key was not revoked and that your account is active.
Your account is not active. Trying again will not help: contact us.
No. The plugin, the website assistant and your calls keep working. The store only stops receiving plugin
updates until you give it a new key.
# Refund requests
Source: https://docs.ruleinside.com/refunds
Refunds your assistant recorded from shoppers. You decide every one, and nothing is refunded automatically.
When a shopper asks your assistant for a refund, the assistant records the request and you decide it under
**Refund requests**. The page says it plainly: "Requests your assistant recorded from shoppers. You decide each one -
nothing is refunded automatically."
## RuleInside never issues a refund
Approving or declining a request only **records your decision**. RuleInside does not move any money and does not
tell the shopper what you decided. A notice above the list reminds you of this:
> **Your decision is recorded here - the refund and the reply are yours to send**
>
> RuleInside records what you decide. It does not issue the refund. When a request comes in we email the shopper to
> confirm we passed it to you - nothing more, and never that a refund is coming. After you decide, refund the order
> in your store as you normally would, and tell the shopper yourself: your decision is yours to deliver.
On the call, your assistant records the request and tells the shopper: "Your refund request has been recorded and
sent to the store owner to review. They will get back to you about it."
RuleInside emails the shopper to confirm the request was passed to you. The email never says a refund is coming.
Open **Refund requests** and click **Approve** or **Decline**.
If you approved it, refund the order in your store as you normally would. Either way, tell the shopper your
decision yourself.
## How a request gets here
* **Only shoppers your store has recognised can make a request**, for example a shopper signed in to your store.
Guests and callers your store has not identified cannot have a request recorded.
* **One open request per order.** If the shopper asks again while it is waiting, their new reason is added to the
same request, marked "(follow-up)".
* **A request can be approved for 72 hours** from when it was recorded. After that it shows as **Expired**. If the
shopper asks again after it expired, the same request reopens for another 72 hours, the new reason is added marked
"(asked again)", and you are notified again.
While requests are waiting, your dashboard shows a banner, for example "A refund request is waiting for you", with
"A shopper asked your assistant for a refund. Nothing is refunded until you decide." and a **Review refund requests**
button that opens this page.
## What each request shows
* "Order" and your store's order number, with an amount. **The amount is the order's total at the time of the
request**, not an amount the shopper chose. It is shown in the order's currency, or as a dash if unknown.
* A status badge, and a red **Expired** badge when a waiting request is past its 72 hours.
* The shopper's name, or their email, or "Customer details unavailable", then the date and time the request was
recorded, in your local time.
* The shopper's reason in their own words, with any "(follow-up)" or "(asked again)" additions.
* Once decided: "Decided by", the email of the person who decided, and when, followed by the decline note if one was
written.
The page lists 50 requests at a time, newest first. When there is more than one page, use **Previous** and **Next**.
At the top, a line tells you how many requests are waiting, for example "2 requests are waiting for you".
## Statuses
| Badge | What it means |
| - | - |
| **Waiting for you** | Not decided yet. |
| **Approved** | You approved it. No money has moved: refund the order in your store. |
| **Declined** | You declined it. |
| **Refunded** | Not used at present. Requests do not reach this status today. |
| **Expired** (red, next to Waiting for you) | Waiting, but past its 72 hours. It can be declined, not approved. |
## Approving and declining
**Approve** asks you to confirm ("Approve this refund request?"), then records the request as approved.
Approve is not offered on an expired request.
**Decline** opens **Decline this refund request**:
* "Note why, in your own words. This is kept with the request for your records - it is not sent to the shopper, so
tell them yourself."
* A text box for your note. It is optional, up to 1,000 characters. The example text reads "For example: this order is
outside our 30-day returns window."
* **Decline request** records the decline, with your note.
**A decision is final.** Once a request is approved or declined, the buttons disappear and it cannot be changed in
the portal. Check the order before you confirm **Approve request**.
Every decision is kept on record: what was decided, who decided it, when, and the note.
**Messages you may see:**
| Message | What it means |
| - | - |
| "Your decision has been recorded." | Done. |
| "The decision could not be recorded. Please try again." | Nothing was recorded. Try again. |
| "That request has already been decided - the list has been refreshed." | Someone else in your team decided it first. |
| "That refund request no longer exists." | The request is gone. |
| "This request is past its deadline, so it can no longer be approved. You can still decline it. If the shopper asks again, it reopens for your approval and you will be notified." | It expired. Decline it, or wait for the shopper to ask again. |
## Who can see and decide
* **Everyone in your workspace** can see the list, including shoppers' names and emails.
* **Deciding** needs the refund-approval permission, which only **owners** have.
* Someone without it sees, under each waiting request: "Deciding refund requests needs the refund-approval
permission, which is not switched on for this workspace yet."
* If the permission could not be checked at that moment, the buttons are hidden and the request says: "We could not
confirm your refund permissions just now, so the decision buttons are hidden. This is a temporary fault on our side,
not a change to your account - please refresh in a moment."
## When the list is not shown
| What you see | What it means |
| - | - |
| "No store is linked to this workspace yet, so there are no refund requests to show." | No store is connected. |
| "We could not load your refund requests" | The list could not be read. This is not the same as having none. Refresh in a moment, and contact support if it keeps happening, so a waiting shopper is not left without an answer. |
| "No refund requests yet." | There are none. |
## Common questions
No. Approving only records your decision. Refund the order in your store, then tell the shopper.
Not from RuleInside. They only received an email saying the request was passed to you. You tell them.
No. It is for your records only.
It expired: more than 72 hours have passed since it was recorded. Decline it, or wait: if the shopper asks again,
it reopens for 72 hours and you are notified.
Only owners can decide refund requests.
It is the full order total at the time of the request, not a partial amount. Refund whatever you decide in your
store.
Not in the portal. A decision is final once recorded.
Only shoppers your store has recognised, for example signed in, can have a request recorded. Guests cannot.
That is expected. Further requests for the same order are added to the open request, marked "(follow-up)" or
"(asked again)".
# Who can do what
Source: https://docs.ruleinside.com/roles
The three roles in your organization, owner, member and primary owner, and exactly what each one can do.
Everyone in your organization holds one of two roles: **Owner** or **Member**. One owner is also the
**Primary Owner**. You see these as badges in the **Role** column of [Members](/members).
| Role | In short |
| - | - |
| **Owner** | Can see and change everything in the portal, including billing and other members' roles. |
| **Member** | Can see every page and invite colleagues, but cannot change how your assistant works. |
| **Primary Owner** | Exactly one person per organization. An owner with a few extra powers: renaming the organization, changing its logo and transferring ownership. |
"Owner" in this guide means **any** owner, not only the primary owner. If you make a colleague an owner, they can
manage billing and every other owner-only control straight away.
## The roles in detail
### Owner
An owner can change everything a member can only see. The organization's creator is the first owner. You can make
more owners with **Update Role** on the [Members](/members) page.
### Member
A member sees the same pages as an owner and can read everything on them. Buttons that would change something
are hidden or greyed out. A member can still:
* invite other people, but only with the **Member** role;
* opt a contact out of calls and messages (a customer's request to stop must never wait for an owner);
* preview an automation, which only shows what would happen and contacts nobody;
* read the refund requests queue (deciding a refund is for owners);
* change their own profile: name, picture, email, password and two-factor sign-in.
### Primary Owner
There is exactly one primary owner. They hold the **Owner** role and carry an extra amber **Primary Owner** badge.
Only the primary owner can:
* save a new organization **name** or **logo** (see [Organization settings](/organization-settings));
* transfer ownership to someone else;
* change the role of, or remove, another owner;
* change or renew an invitation for the **Owner** role.
The primary owner cannot leave the organization or be removed from it, and cannot delete their own login while
they are primary owner. Transfer ownership first (see [Members](/members#transferring-ownership)).
## What only an owner can change
Anything that changes how your assistant talks to customers, contacts them, spends minutes or connects to other
systems:
* your assistants, their knowledge and their tools;
* campaigns, including starting, pausing and stopping them;
* contacts (adding, importing, editing, columns) and their consent;
* calling hours, calling days and the calling consent confirmation;
* phone numbers and WhatsApp, including message templates and sending a message;
* connections to your store and to other systems, the WordPress plugin licence key, and developer settings;
* the website widget;
* automations and deciding refund requests;
* billing: choosing a plan, invoices, the billing portal and cancelling;
* the roles of other members, and pending invitations.
## Everything at a glance
| Action | Primary Owner | Owner | Member |
| - | - | - | - |
| See every page | Yes | Yes | Yes |
| Change assistants, campaigns, contacts, connections and the other areas above | Yes | Yes | No |
| Opt out a contact, preview an automation, read refund requests | Yes | Yes | Yes |
| **Invite Members** | As Owner or Member | As Owner or Member | Only as Member |
| **Update Role** | Anyone except themselves | Only people with the Member role | No |
| **Remove from Account** | Anyone except themselves | Only people with the Member role | No |
| **Transfer Ownership** | Yes | No | No |
| Update, renew or remove a pending invitation | Yes | Yes, for Member invitations | No |
| Save the organization name or logo | Yes | No | No |
| **Leave Organization** | No, transfer ownership first | Yes | Yes |
| Their own profile | Yes | Yes | Yes |
| Billing | Yes | Yes | No, they see **You cannot manage billing** |
## Common questions
On [Members](/members), open the **...** menu on their row, choose **Update Role** and pick **Owner**. The
primary owner can do this for anyone; any other owner can do it for a member. The change takes effect at once,
including access to billing.
Look for the amber **Primary Owner** badge in the **Role** column on the [Members](/members) page.
They are a member. Billing is for owners only. Make them an owner, or download the invoices yourself.
No. There is no limit on how many people you invite, and no plan is billed per person. Adding or removing
members never changes your bill.
No. Only the primary owner can save the organization name or logo. Anyone else who tries sees "Could not update
Organization. Please try again."
Invite colleagues as members when they need to follow what your assistant is doing. Keep the owner role for the
people who decide how it contacts your customers and who manage billing.
# Testing before you go live
Source: https://docs.ruleinside.com/testing
Hear your agent, try your website assistant, call your own number, place a test call and send yourself a WhatsApp message, then check everything before customers reach it.
When you finish this page you will have talked to your agent the way your customers will, on every channel you use,
read back what happened, and ticked off the checks that matter before real customers are called or messaged. We
suggest you set aside about an hour for a first full pass.
Some tests are real. Calling your own number, a campaign **Send test call** and a WhatsApp send are real calls and
messages, and a test call is billed like any other call. Only ever test with your own phones.
## The ways to test
| Test | Where | Who can use it | What to know |
| - | - | - | - |
| **Preview** | **Agents**, in the agent's editor header | Owners | Talks to the real agent, by voice or text, with its **saved** settings. It treats you as a guest. |
| **Live preview** | **Website widgets**, in the widget editor | Everyone on your team (only owners can change the settings) | The real widget, with a real conversation. It counts towards the daily conversation limit. |
| Calling your own number | Your phone | Anyone | A real phone call, answered by the number's **Inbound agent**. |
| **Send test call** | A draft phone campaign | Owners | Real and billable. Calls only the numbers in **Test numbers**. |
| **WhatsApp send** | **WhatsApp send** | Owners | Sends one real template message to one saved contact. |
## Talk to your agent with Preview
**Preview** is the quickest test: it runs in your browser and needs no phone.
Preview uses the agent's **saved** settings. If you changed something, click **Save changes** in the bar at the
bottom of the editor before you start.
Open **Agents** and click **Preview** in the editor header. Only owners see the button.
Click **Text** or **Voice**, then **Start text chat** or **Start voice call**. For voice, your browser asks to
use your microphone. The status moves from "Not connected" to "Connecting…" and "Connected".
Ask what your customers ask. In voice, just talk: the panel shows "Agent is speaking…" or "Listening, just talk".
Click **End**, or close the panel.
What Preview does and does not show:
* It is the real agent: its greeting, voice, instructions, knowledge and tools.
* It does **not** identify you as a customer. Store tools answer as they would for a guest: an order's details are
given only when you say its order number together with the billing email and postcode on that order. Test
order look-up for a signed-in customer on your website instead (see below).
* We recommend testing in **Voice** at least once. Text shows what the agent says; only voice shows how it sounds,
how quickly it answers and how it handles interruptions.
If it fails you see "Could not start the preview.", "Could not connect.", "Preview error." or "Conversation ended:"
with a reason. See [Testing an agent with Preview](/managing-agents#testing-an-agent-with-preview).
## Test the website assistant
### In the widget editor
Open **Website widgets** and click the pencil on a widget. The **Live preview** on the right is the real widget,
updated a moment after each change you make in the editor.
* It is a real conversation with your agent. Your browser asks for the microphone, and messages you type are really
sent.
* Like a conversation on your website, it needs the widget switched on, your account active and on a paid plan, and
it counts towards your daily conversation limit.
* **Width (px)** and **Height (px)** are not shown in the preview.
* The preview shows your changes before you save, but your website does not change until you click **Save changes**.
### On a copy of your site
If you have a staging copy of your store, we recommend trying the widget there before your live site:
1. Add the staging site's address, starting with `https://`, on its own line in **Allowed domains**, and click
**Save changes**. The widget only appears on the addresses listed there. (`http://` is accepted only for
`localhost` addresses on your own computer.)
2. Paste the snippet into the staging site or, on WordPress with the RuleInside plugin, paste the **embed id** into
the plugin's **Widget ID** field (WordPress admin, menu **RuleInside**).
3. Open the staging site as a shopper would and start a conversation.
The same snippet works on every address in **Allowed domains**. See [Website assistant](/website-assistant) and
[WordPress plugin](/plugin).
### As a signed-in customer and as a guest
Your assistant gives order details to a customer logged in to your store, about their own orders. Someone not
logged in gets an order's details only by giving its order number together with the billing email and postcode on
that order; otherwise, general help. Try both:
* **Signed in:** log in to your store as a customer who has an order (a test account of your own is ideal) and ask
about that order. In a voice conversation, a signed-in shopper your store recognises is greeted with the
**Returning-shopper greeting**, as long as it mentions both AI and recording.
* **As a guest:** log out, ask the same question without giving the billing email and postcode, and check that no
order details are given.
Order look-up needs your store connected under [Connections](/connections).
## Call your own phone number
Open **Phone numbers**. The number's **Inbound agent** must show an agent: a dash means no agent answers calls to
it. Press **Assign** to choose one.
Listen to the greeting from start to end.
We suggest you interrupt the agent, say "ok" while it talks, spell an order number or an email address, stay
silent for a while, and, if transfers are set up, ask to speak to a person.
Open **Calls**, choose the agent in the **Agent** bar, and look for a **Phone inbound** call from your number.
Its summary and outcome appear a short time after the call ends: refresh if they are missing.
What you hear on a call is shaped by [Turn taking](/agent-settings#turn-taking), **Call limits** and **Conversation
behaviour** in [Agent settings](/agent-settings).
## Place a campaign test call
A campaign can call your own phones before it calls anyone else. Test calls are for phone campaigns only.
Click **New campaign**, choose **Phone call**, and fill in the campaign as you mean to run it. Creating it only
saves a draft: nobody is called.
In **Test numbers (optional)**, type your own numbers, one per line, in international format such as
`+15551234567`. Up to 20.
Open the campaign and click **Send test call**, then **Place test call**. The campaign stays a draft.
A test call is a real, billable call, and it counts towards your daily call limit. It **skips the consent check**,
because a test number is not one of your contacts. Never put a customer's number in **Test numbers**.
A test call is still checked against your store's calling consent confirmation, your do-not-call list and opt-outs,
the campaign's calling hours and days, and your daily call limit. You then see "Test call placed to \{count}
number(s)." or "No test call was placed." with the reason. A common one: you are testing outside the campaign's
calling hours. See [Send test call](/campaign-results#send-test-call).
## Send yourself a WhatsApp message
WhatsApp campaigns have no test call. To test a template, send it to your own number from **WhatsApp send**. Before
you start, check that:
* your WhatsApp number is connected on **Phone numbers** (for a number our team connected for you, with
**Messaging enabled** on and an agent assigned);
* the template is registered under [WhatsApp templates](/whatsapp-templates) with **Kind** set to **Message** and
**Enabled** on;
* **you** are a saved contact, with your number in international format starting with `+`, and WhatsApp consent is
recorded on your contact with **Consent** (see [Contacts](/contacts#recording-consent-for-one-contact));
* an owner has confirmed **Settings › Calling consent**.
On **WhatsApp send**, type your number in **To (phone number)**, choose the **Template**, and fill in any
**Template variables**.
Click **Send message**, then **Send now** in the confirmation.
Look at your row in **Recent sends**. **Status** shows where the message is; **Delivery** shows what WhatsApp
reported, for a number connected through Meta.
* A **Marketing** template is sent only inside the calling hours where you are. Outside them it waits as **Queued**.
* A number that is not a saved contact is refused, and the send log shows **Suppressed** with `opt_out`.
* If you reply **STOP** from your test phone, you opt yourself out of WhatsApp messages and calls from your store.
Reply **START** to opt back in to WhatsApp messages (calls stay stopped).
See [Sending a WhatsApp message](/whatsapp-send).
## Pre-launch checklist
Work through this list before real customers can reach your agent.
| Check | Where to look |
| - | - |
| The greeting says the caller is talking to an AI assistant and that the call may be recorded. The portal does not stop you saving a greeting without them. | **Agents**, **First message** and **Recording notice**. On the website, also the **Returning-shopper greeting**. |
| The instructions sound natural read aloud: no lists, links or symbols the agent would read out awkwardly. | **System prompt** in the agent editor. See [Writing instructions](/writing-instructions). |
| The agent answers your five most common customer questions correctly from your knowledge. | **Preview**, then [Knowledge](/knowledge). |
| Order look-up works for a signed-in customer, and gives a guest no order details without the order number, billing email and postcode. | Your website, signed in and signed out. |
| Every transfer number is answered by a person. We suggest calling each one yourself. | **Transfer to a person** in the agent editor. |
| Calling consent shows **Confirmed**. Until it does, no calls or WhatsApp messages are sent. | **Settings › Calling consent**. |
| Calling hours and days are right, and **Timezone for customers we cannot place** is set the way you want. | **Settings › Calling consent**, **Calling hours and days**. |
| The contacts you plan to call show **Calls: callable**. | [Contacts](/contacts). |
| A test call was placed and you listened to it. | Your own number, or **Send test call**. |
| The test shows on the **Calls** page, with a summary. | [Calls](/conversations), with the right agent selected. |
## Reading a test conversation
Open **Calls**, choose the agent, and click the date of your test call. The call's page shows, from top to bottom:
* **Recording**: listen to the greeting and the agent's voice as a customer heard them.
* **Details**: the channel, the duration and the **Result** (**Success** or **Failure**).
* **Summary** and **Data collected**: what the agent understood and captured, for example a lead's email.
* **Evaluation**: each check your agent runs, with its result and explanation, when your agent has any.
* **Transcript**: the whole conversation, message by message.
In the transcript, look for:
* **Slow answers.** The figure after an agent message's time, such as "· 1.41s", is how long the agent took to
respond. If answers feel slow, see [Turn taking](/agent-settings#turn-taking).
* **Interruptions.** **interrupted** marks a message that was cut off. If the agent stops when the caller says
"ok", add those words to **Words that do not interrupt**.
* **Actions.** Rows with a wrench icon are actions the agent took, such as looking up an order, with "ok" or "error".
Open one to see **Request** and **Response**.
* **Knowledge look-ups.** A row with a book icon shows the agent checked your knowledge. If it never checks a
document you expected, see [How the agent uses a document](/knowledge#how-the-agent-uses-a-document).
See [Calls](/conversations) for every card and notice.
## A suggested way to roll out
You do not have to switch everything on at once. A careful order, which we recommend:
1. **Website first.** Try the widget on a staging copy, then add your live address to **Allowed domains**. To stop
it at any time, switch **Widget enabled** off and click **Save changes**: your snippet does not need to change.
2. **Your phone line.** Assign the **Inbound agent** and read the **Calls** page every day for the first week.
3. **A small campaign.** Upload a short list of people who agreed to be called, launch, and read the launch result
and every call before you reach a larger group. An uploaded list holds at most 1,000 rows, and a campaign
reaches at most 5,000 contacts.
4. **Automations last.** Use **Preview** on a rule to see who it would reach (it never places a call), and keep one
rule switched on at a time.
5. **WhatsApp.** Send a template to yourself, then to a small group.
## Common questions
Preview uses the saved settings. Click **Save changes**, then start a new Preview.
Only account owners see it. Ask an owner of your team.
Preview treats you as a guest, so it finds an order only when you give its order number together with the
billing email and postcode on that order. To test a signed-in customer, use your website while signed in to your
store as a customer.
No. It is a real call, billed like any other, and it counts towards your daily call limit.
Read the reason in the result. The most common: your store's calling consent is not confirmed, you are outside
the campaign's calling hours, a test number is not in international format, or the daily limit is reached.
No. Test calls are for phone campaigns only. Send the template to your own number from **WhatsApp send**.
Yes. It is a real conversation and counts towards your daily conversation limit.
No. A test call skips the consent check, so only use your own phones.
## Next steps
Find a problem by what you see, and how to fix it.
Every card on a call's page, explained.
Change what you heard in your tests.
Launch, pause and read your first campaign.
# Tools
Source: https://docs.ruleinside.com/tools
What your agent can do during a conversation, how to add your own tools, and the built-in system tools.
Open **Tools** to see what the agent chosen in the **Agent** selector can do during a conversation: RuleInside's
store tools, your own custom tools and the built-in system tools. Everything on this page applies to that agent; to
work on another agent, choose it in the selector.
## Who can change tools
Any member can see this page. To change anything you need **both**:
* to be an **owner** of the account, and
* the active agent to be at **Full control** (see [Edit levels](/agent-settings#edit-levels)). Agents you create
are always at Full control; your main agent often is not.
Otherwise the page is read-only: there is no **Add tool** button, no **Attach**, **Remove**, edit or delete
buttons, and the switches are greyed out. Transfer numbers follow a different rule, described under
[Transfer rows](#transfer-rows).
## The tools list
The page title reads "What … can do during a conversation." with your agent's name. Above the list:
* **Add tool** (only when you can change tools).
* A search box, **Search tools…**, which matches names and descriptions.
* Filters: **All**, **Webhook**, **In-page**, **Integration**. They also apply to RuleInside's tools.
The search and filters appear once there is at least one tool. The list shows, in order:
1. **RuleInside's store tools** on the agent (such as product search, order look-up and coupon check), each with a
padlock and **Managed**. You cannot change or remove them, and they always stay on the agent.
2. **Your custom tools attached to this agent**, with their name, type, description and URL (for Webhook and
Integration tools), and the buttons **Remove**, edit (pencil) and delete (trash).
3. **Available to attach** (only when you can change tools): your custom tools that are not on this agent, each
with **Attach**, edit and delete.
The footer reads: "Built-in store tools are managed by RuleInside. Custom tools are reusable, attaching one here
doesn't remove it from other agents."
Each tool shows a type badge: **Webhook**, **In-page**, **Integration**, **Built-in** or **Tool**. The same Webhook
tools are labelled **Store** in the agent editor's Tools card.
## Buttons on a custom tool
| Button | What happens | Can it be undone? |
| - | - | - |
| **Attach** | Adds the tool to this agent. Saves straight away, with no Save button. | Yes, with **Remove**. |
| **Remove** | Takes the tool off this agent only. The tool itself is kept and stays on your other agents. Saves straight away. | Yes, with **Attach**. |
| Pencil (**Edit**) | Opens the tool's settings in the **Edit tool** dialog. Not offered for Integration tools. | Edits change the tool on **every** agent it is attached to. |
| Trash (**Delete**) | Deletes the tool for good, from every agent. | **No.** |
Deleting a tool asks you to confirm ("Delete this tool?") and cannot be undone. To use the tool
again you must create it again, including any secret value.
Messages you may see:
* Attach or Remove failed: "Could not update tools." (or the reason), and the page reloads.
* On your main agent below Full control: "This agent is managed, custom tools can only be attached to your own
agents."
* Delete: "Tool deleted." or "Could not delete the tool."
* Edit could not open: "Could not load the tool."
## Adding or editing a tool
Click **Add tool** to open **New tool** ("Give your agent a new capability."), or the pencil on a tool to open
**Edit tool** ("Change this tool. The type can't be changed."). Choose a type:
* **Webhook** ("Call your API"): the agent calls an address on your server.
* **In-page** ("Run in the widget"): the tool runs inside your website assistant.
* **Integration** ("MCP server"): connects an MCP server.
In **Edit tool** the type is locked.
### Fields for every type
| Field | What it does | Values and limits |
| - | - | - |
| **Name** | The name the agent uses to call the tool (placeholder "Check loyalty"). "The agent calls the tool by this name. Spaces become underscores." | Converted as you type: lower case, any character other than a letter or number becomes an underscore, must start with a letter, up to 64 characters. When the result differs from what you typed, the hint shows "Saved as:" and the converted name. At least 2 letters or numbers. |
| **Description** | "When and how the agent should use this tool." The agent reads this to decide when to use the tool. | Up to 1,000 characters |
A name reserved by RuleInside is refused. A tool may not refer to RuleInside's reserved identity or system
variables (such as names starting with `system__`) in its name, description, URL, header or parameters; the
message says "Tools may not reference reserved identity or system variables".
### Webhook fields
| Field | What it does | Values and limits | Default |
| - | - | - | - |
| **URL** | The address the agent calls (placeholder "[https://api.yourstore.com/endpoint](https://api.yourstore.com/endpoint)") | "Must be a public URL, internal/private addresses are blocked." Must start with `https://`, use a public domain name (not an IP address or `localhost`), and contain no user name or password. | `https://` |
| **Method** | How the address is called | GET, POST, PUT, PATCH, DELETE. With GET the parameters go in the address; otherwise in the request body. | POST |
| **Parameters** | The inputs the agent sends | See below | None |
| **Authentication (optional)** | A header sent with every call. "Stored securely as a workspace secret, never shown again." | Header name (placeholder "Header, e.g. X-Api-Key": letters, numbers and hyphens, up to 64) and value (placeholder "Secret value"). Both are needed, otherwise no header is sent. | None |
**Parameters.** Click **+ Add** to add a row ("No parameters. Add the inputs the agent should provide." while
there are none). Each row has:
* a name box (placeholder "param\_name"), converted like the tool name;
* a type: string, number, integer or boolean;
* **from variable (optional)**: when filled, the value comes from that conversation variable instead of from the
agent, and **req** is greyed out. RuleInside's reserved variable names are refused;
* **req**: the agent must supply this value;
* a trash button to remove the row.
Parameters have no description box: the parameter's name is used as its description. Give each parameter a clear
name, such as `order_number` or `customer_email`.
**Authentication when editing.** If the tool already has a value, the box reads "Leave blank to keep current key".
Leave it blank to keep the stored value, or type a new one to replace it. You cannot read the stored value back,
and you cannot remove authentication from an existing tool in the portal: to drop it, delete the tool and create it
again without it.
### In-page fields
**Parameters**, as above. "In-page tools run inside the embedded widget and need a matching handler on your site
to do anything. This is an advanced option."
### Integration fields
| Field | What it does | Values | Default |
| - | - | - | - |
| **Server URL** | The MCP server's address (placeholder "[https://mcp.yourservice.com](https://mcp.yourservice.com)") | Same rules as the Webhook URL | `https://` |
| **Transport** | The connection type the server uses | **Streamable HTTP** or SSE | Streamable HTTP |
| **Secret token (optional)** | A token sent to the server (placeholder "Bearer token") | Text | Empty |
"Every tool call requires approval by default. MCP integrations may require an upgraded plan."
Integration tools must be switched on for your account by RuleInside; they are off unless we have enabled them.
An Integration tool cannot be edited after it is created: delete it and create it again.
### Saving the dialog
Click **Create tool** ("Creating…") or, when editing, **Save changes** ("Saving…"). **Cancel** closes the dialog
without saving.
* A new tool is created and attached to the active agent in one step: "Tool created and attached." If the tool is
created but cannot be attached, it appears under **Available to attach**; click **Attach** there.
* An edit shows "Tool updated." and changes the tool on every agent it is attached to.
* Failures show the reason, or "Could not create the tool." / "Could not update the tool."
## System tools
The **System tools** panel on the right reads "Allow the agent to perform built-in actions." and shows how many
are active, including the transfer tools when they are on.
Five switches: **End conversation**, **Detect language**, **Skip turn**, **Play keypad touch tone** and
**Voicemail detection**.
* Each switch saves as soon as you click it: "Updated." or "Could not update." (the page then reloads).
* Only an owner can use them, and only on an agent at **Full control**. Otherwise they are greyed out with "This
agent is managed, its system tools are configured by RuleInside." or "Only account owners can change these, ask
an owner of your team."
* **Voicemail detection** also decides what **When a call reaches voicemail** shows under
[Call limits](/agent-settings#call-limits).
### Transfer rows
* **Transfer to agent**: always locked. Only RuleInside sets it up.
* **Transfer to number**: shows a settings icon when the agent has transfer numbers. When the agent editor shows
the **Transfer to a person** card, the row links to it, with an arrow if you can change transfers there (an owner,
with the agent at **Instructions** or above) or a padlock if you cannot. "Transfer numbers are set on each agent,
under "Transfer to a person"." See [Transfer to a person](/agent-settings#transfer-to-a-person).
## What you might see
| Message | What it means |
| - | - |
| "Tool builder isn't enabled on your plan. Contact RuleInside if you'd like this feature enabled for your account." | Tools are not switched on for your account. Contact RuleInside. |
| "This workspace is not linked to a store yet." | Your workspace is not connected to your store yet. |
| "Couldn't load your tools. Refresh to try again." | The page failed to load. Refresh it. |
| "No tools yet" | The agent has no tools. Owners see "Add a tool to let this agent call your API."; otherwise "This agent has no tools attached." |
| "No matching tools on this agent." | Nothing matches your search or filter. |
| "No tools on this agent yet." | You have custom tools, but none is attached to this agent. |
Row buttons are greyed out while an attach, delete or load is running.
## Common questions
Either the active agent is not at **Full control** or you are not an owner. Your main agent is often below Full
control; agents you create are always at Full control.
No. RuleInside's store tools are marked **Managed** and always stay on the agent.
There is no undo. Create the tool again, including any authentication value, and attach it to each agent that
used it.
It must start with `https://`, use a public domain name (not an IP address or `localhost`), and contain no user
name or password.
Yes. Choose the other agent in the selector and click **Attach** on the tool. Editing the tool later changes it
on every agent.
**Remove** takes the tool off this agent only and keeps it for your other agents. Delete (the trash button)
erases it from every agent, straight away and for good.
Nowhere. Secret values are never shown again. To change it, edit the tool and type a new value.
In the agent editor, under **Transfer to a person**. The **Transfer to number** row on this page links there.
No. Delete it and create it again.
If you see "Could not update.", the change was not saved and the page reloaded with the current value. Check
that you are an owner and that the agent is at **Full control**.
# Troubleshooting
Source: https://docs.ruleinside.com/troubleshooting
Find your problem by what you see, check the likely causes in order, and look up any message the portal shows you.
This page helps you fix a problem yourself in a few minutes. Find the section for the part of RuleInside that is
not doing what you expect, open the symptom that matches yours, and work through the causes in order: the most
likely comes first. If the portal showed you a message, jump to [Messages you may see](#messages-you-may-see).
Three things are worth checking first, because each affects many pages: whether you are an **owner** (members can
see everything but change little), whether you clicked the save button, and whether **Settings › Calling
consent** shows **Confirmed**.
## Find your problem
### Your agent
1. You are a **member**: only owners can change agents. Ask an owner of your team.
2. The agent's **edit level** does not include that setting. Your main agent is often below **Full control**.
3. The agent is **View only**: a banner says so, and nothing can be changed.
See [Edit levels](/agent-settings#edit-levels).
1. Closing a side panel does not save it. Click **Save changes** in the bar at the bottom of the editor.
2. The save was refused: look for "Some settings need fixing before this can be saved." and open the highlighted
panel.
3. **Preview** uses saved settings only: save, then start a new Preview.
1. The customer is not logged in to your store and did not give the order number together with the billing
email and postcode on that order. Without one or the other, no order details are given.
2. You are testing in **Preview**, which always treats you as a guest.
3. Your store is not connected: check [Connections](/connections).
1. The document is in **Auto** mode but RAG was never saved: open **Configure RAG**, check **Enable RAG** is on,
and click **Save**.
2. The document is under about 500 bytes: keep it in **Prompt** mode.
3. It is attached to a different agent: check the **Agent** selector on **Knowledge**.
See [Knowledge](/knowledge).
1. You are a member: these are for owners.
2. On **Tools** and **Knowledge**, the agent is below **Full control**. Agents you create are always at Full
control.
3. Tools or knowledge are not switched on for your account: the page says so. Contact us.
Your storefront did not pass in a value for that variable, so its default was used. Check the default under
**Personalization variables**. See [Agent settings](/agent-settings#personalization-variables).
### Website assistant
Check in this order:
1. The exact address of your site, starting with `https://`, is in **Allowed domains**.
2. **Widget enabled** is on, and you clicked **Save changes**.
3. The embed id was not rotated after you pasted the snippet. If it was, paste the new snippet, or the new
**Widget ID** in the WordPress plugin.
4. Your account is active and on a paid plan.
5. Today's conversation limit is not reached: look for a notification in your portal.
See [Website assistant](/website-assistant).
1. You did not click **Save changes**.
2. You changed **Widget mode**, **Width (px)** or **Height (px)**: those change the snippet itself. Copy it again
and paste it onto your site.
3. You cleared a field: clearing does not reset it today. Type the value you want instead.
1. More than 30 conversations started on this widget in one minute. It clears when the minute is over.
2. Your account reached today's conversation limit: look for a notification in your portal. It clears at
midnight UTC.
1. It does not say both that the shopper is talking to an AI and that the call may be recorded.
2. The shopper is a guest, or used text: the greeting is only for signed-in shoppers your store recognises, in
voice.
1. You are a member: only owners can.
2. You are not on a paid plan, or your account is not active.
3. Every agent already has a widget: each agent can have one.
Check the licence key's **Last used** on **Settings**, **Plugin**. **Never used** means it was never pasted into
the plugin, or pasted wrongly. Also check it was not revoked and your account is active. See
[WordPress plugin](/plugin).
### Phone calls
1. The number's **Inbound agent** on **Phone numbers** shows a dash: no agent answers it. Press **Assign** and
choose one.
2. Your subscription is not active (**Unpaid**, **Canceled**, **Incomplete**, **Expired** or **Paused** on
**Billing**): your assistant is paused.
3. On a pay-as-you-go bundle, a declined card can stop everything, inbound calls included.
4. The number was released. Import it again.
1. You are a member: only owners can.
2. **Label** is empty: **Import** stays greyed out until it is filled.
3. The number is not in international format, such as `+15551234567`.
4. The number is already linked to another RuleInside account: contact us.
See [Phone numbers](/phone-numbers).
Check **Maximum call length (minutes)** and **Hang up when the caller stays silent** under **Call limits**.
For slow answers, open **Turn taking**: set **Response eagerness** to Eager and try a shorter **Wait for a reply
(seconds)**. For 'ok', add the words to **Words that do not interrupt** in **Conversation behaviour**.
1. **Transfer to a person** is not switched on for your account: you can remove a transfer but not add one.
Contact us.
2. The number is refused: it must be a mobile, landline or toll-free number in the same country as your store's
phone number, and not your store's own number.
See [Transfer to a person](/agent-settings#transfer-to-a-person).
1. The Calls page shows one agent at a time: choose the right one in the **Agent** bar.
2. The summary and outcome appear a short time after the call ends: refresh.
3. Search covers only the current page: use **Older**.
See [Calls](/conversations).
No recording exists for the call, the caller objected to being recorded (the recording was then deleted), or
there was a temporary problem. Try again in a moment.
### WhatsApp
1. You are not the account owner.
2. A number is already connected and working: the button is hidden on purpose.
3. Self-connection is not available for your account: the card says so. Contact us and we connect it for you.
**At WhatsApp** shows **Rejected** when Meta rejected the template. Fix it in Meta's WhatsApp Manager and get it
approved. Until then, disable it under **WhatsApp templates**: a rejected template that is still **Enabled** is
still tried, and the send shows **Failed**. See [Status at WhatsApp](/whatsapp-templates#status-at-whatsapp).
The name or language code does not match Meta exactly (`en` is not `en_US`), or the template was approved on a
different WhatsApp account. Edit the row to fix it.
Read the reason in **Detail**:
1. `no_consent`: no WhatsApp consent on the contact. Consent to calls does not count.
2. `opt_out`: they opted out, **or** no saved contact has this exact number.
3. `frequency_cap`: they already had 4 marketing messages in 7 days.
4. `no_store_attestation`: **Settings › Calling consent** is not confirmed.
5. `marketing_stopped`: a Marketing template, and the person told WhatsApp to stop marketing messages from your
business. Utility and Authentication templates still reach them.
See [Why a send is suppressed](/whatsapp-send#why-a-send-is-suppressed).
1. A **Marketing** template waits for the person's calling hours. Nothing to do.
2. The contact has no time zone: set one. A Marketing message still waiting for one 7 days after it was queued
is not sent (**Failed**); set the time zone and send it again.
3. WhatsApp is blocking your number, often for a missing payment method in Meta. **Phone numbers** shows what
Meta reports.
**Sent** means WhatsApp accepted it. Look at **Delivery**: **Not delivered** shows WhatsApp's reason in
**Detail**. **No report from WhatsApp** is normal for a number not connected through Meta.
**Send message** stays grey until a number, a template and every variable are filled. Only templates with
**Kind** set to **Message** and **Enabled** on are listed.
A variable came out empty, usually because the contact is missing a detail such as a name. Fill it in on the
contact, or change the variable on the template.
### Campaigns
1. You created a draft but never pressed **Launch campaign**. A schedule does not launch it either.
2. Read the launch result: each line gives a reason, such as **no recorded lawful basis to call** or **outside
local calling hours**.
3. Your store's calling consent is not confirmed.
If nobody was called, nobody is waiting and nobody is queued to be tried again, the campaign is still a draft:
fix the reason and launch again. See
[The launch result](/campaign-results#the-launch-result).
People are probably **Waiting for their calling time**: the summary shows how many and when the first is due.
The launch result names the reason: opted out, on the do-not-call list, no lawful basis, outside calling hours, no
time zone, called too recently, or the daily limit. People left out because of the daily limit or a processing
error are tried again by the same campaign over about a day, up to one day of your daily cap per launch; the
launch result says how many. People left out for any other
reason are not called later by that campaign: create a new campaign once the reason has passed.
None of your numbers can place outbound calls (on **Phone numbers**, **Assign** shows "This number does not
support outbound calls." for such a number). Import a number that can, or contact us.
The result gives the reason: calling consent not confirmed, outside the campaign's calling hours, a number not in
international format or on your do-not-call list, or the daily limit reached.
Calls already under way cannot be recalled. If the message said some batches could not be confirmed stopped,
press **Pause again**.
That is expected: it does not change to **Completed** on its own. Check each message in **Recent sends** on
**WhatsApp send**.
### Contacts and imports
1. A plain import adds people as **Calls: no consent**. Use **Import with consent**, or **Consent** on each
contact.
2. **Settings › Calling consent** is not confirmed.
3. Your calling hours or days exclude now.
The row had no phone number, or its email already belongs to another contact. With **Import with consent**, also
any number without a country code ("invalid number").
Only owners can import; members see a note instead. For an owner, both buttons stay greyed out until a store is
linked to your workspace, and **Import with consent** also when the list of lawful bases could not load: refresh.
Search covers only the current page of 50 contacts. Use **Next** and search again.
An opt-out cannot be undone from the portal, by re-importing or by recording consent again. On WhatsApp only, the
customer can opt back in by replying START.
The column was removed under **Manage fields**: while it is removed, its values are not sent to your agent on
outbound calls. Add it back with **Add column**. The values are still stored.
### Store connection
You have not approved the connection in your store yet. The approval page is valid for 10 minutes: start again
if it expired.
Read the line under the status on the **Connect your store** card. Common causes: you approved it on a
different website from **Your store address**, you approved less than read and write access, or your store was
offline. See
[Connections](/connections#what-the-last-attempt-says).
For now our team connects FluentCart stores. Contact us.
Check the connection is **Active**, the header is exactly `X-RI-Ingest-Secret`, the secret is the latest one,
and the whole URL was copied.
### Team and access
1. You are a member: you can invite only as **Member**.
2. One address in the form already belongs to a member: the whole form is refused. Take it out.
3. The address already has a pending invitation: renew or remove it first.
See [Members](/members).
Ask them to check their junk folder. Otherwise remove the invitation and invite them again: **Renew Invitation**
does not send an email.
It is more than 7 days old, was removed or already used, or they signed in with a different email. An owner can
**Renew Invitation** on an expired one.
They are a member. Members can see every page but not change how your assistant works. An owner can use
**Update Role**. See [Who can do what](/roles).
Only the primary owner can rename the organization. The primary owner cannot delete their login or leave until
they transfer ownership on [Members](/members#transferring-ownership).
If your codes are in a password manager that syncs, use it on another device. Otherwise, contact us.
### Billing
They are a member: billing is for owners. Make them an owner, or download the invoices yourself.
1. On a pay-as-you-go bundle, a declined card. Update it with **Visit Billing Portal**.
2. Your plan's badge on **Billing** is **Unpaid**, **Canceled**, **Incomplete**, **Expired** or **Paused**.
1. Your daily call limit is reached: it clears within 24 hours.
2. On a pay-as-you-go bundle, outbound calls and messages can stop for the rest of the period after the second
extra bundle.
See [Plans and limits](/limits).
Look at the **Minutes** card on the **Billing** page. Calls from the last day may not be counted yet. If the card
is not shown, reload the page later, or contact us.
## Messages you may see
Find the message you see, then follow **What to do**. Text in curly braces, such as \{count}, stands for a
number the portal fills in.
| Message | What it means | What to do |
| - | - | - |
| **Your agent** | | |
| "Only account owners can change agents, ask an owner of your team to make edits." | You are a member. | Ask an owner. |
| "Permission denied: this action requires an account owner (billing) permission." | A member's change reached the server and was refused. | Ask an owner. |
| "Your plan is view-only. Contact us to change your agent, or upgrade for self-service editing." | This agent is **View only**. | Contact us. |
| "Some settings need fixing before this can be saved. Open the highlighted section to see what." | A field is empty or out of range. Nothing was saved. | Open the highlighted panel and fix the field. |
| "Pick a different language, or contact support to change the voice model for you." | The language does not match the speech engine. | Choose another language, or contact us. |
| "Transfer to a person is not enabled for this account. You can remove a transfer, but not add or change one." | Transfers are off for your account. | Contact us to add a transfer. |
| "Transfers need a phone number on this store first." | Your account has no active phone number. | Import a number on **Phone numbers**. |
| "This agent is managed, documents can only be attached to your own agents." | Your main agent is below **Full control**. | Use an agent you created. |
| "Tool builder isn't enabled on your plan. Contact RuleInside if you'd like this feature enabled for your account." | Tools are off for your account. | Contact us. |
| **Website assistant** | | |
| "The website widget is available on paid plans." | You are not on a paid plan. | Choose a paid plan. |
| "A widget already exists for this agent." | Each agent has one widget. | Open that widget instead. |
| "This agent is not fully provisioned yet. Try again shortly." | The agent is still being set up. | Wait a few minutes and try again. |
| "Enter a full origin like [https://shop.example.com](https://shop.example.com)." | An **Allowed domains** line is not a site address. | Type only `https://` and the site's address. |
| "Use a 6-digit hex color like #1a2b3c." | A colour is not in the right form. | Type a colour such as `#1a2b3c`. |
| "Your website assistant is paused. Contact support to turn it back on." | Your website assistant was paused. | Contact us. |
| "Plugin downloads are paused for this account" | Your account is not active. | Contact us. |
| **Phone calls** | | |
| "That phone number is already linked to another client." | The number is on another RuleInside account. | Contact us. |
| "Could not verify phone-number ownership right now. Please retry." | A temporary check failed. | Try again in a moment. |
| "Could not import the phone number." | The import did not complete. | Check the number, Account SID and Auth Token, then try again. |
| "This number does not support outbound calls." | The number cannot place calls. | Use another number for campaigns. |
| "Recording deleted at the caller’s request" | The caller objected to being recorded. | Nothing: the call keeps its outcome and duration. |
| "No calls yet." | This agent has no calls. | Check the **Agent** bar shows the right agent. |
| **Calling consent** | | |
| "Outbound calling is paused" | Calling consent is not confirmed. No calls or WhatsApp messages are sent. | An owner presses **Review and confirm**. |
| "Please re-confirm" | The wording changed. Your calls are still running. | Read it and confirm again when you can. |
| "Only an owner of this workspace can change this." | You are a member. | Ask an owner. |
| "Your account has no country set, so we cannot tell which rules apply. Choose one below and confirm." | No country is set. | Choose the country and confirm. |
| "We could not check your confirmation, so nothing is being sent. This is on us. Please try again shortly." | The confirmation could not be read. | Try again shortly. |
| **Campaigns** | | |
| "You have no outbound phone number yet. Add one on the Phone numbers page, then create a campaign." | None of your numbers can place calls. | Import a number that can on **Phone numbers**, or contact us. |
| "The selected phone number is no longer available for outbound calling." | The number changed since the draft. | Edit the draft and choose another number. |
| "This account is not active; outbound calling is disabled." | Your account is paused. | Check **Billing**. Pausing and cancelling still work. |
| "No test call was placed." | A check stopped the test call. | Read the reason that follows it. |
| "Launch complete: nobody was dialed" | Nobody could be called right now, and nobody is waiting or queued. | Read the reasons; the campaign is still a draft. |
| "\{count} will be tried again by this campaign when they can next be called" | Your daily call limit was reached, or a check could not be completed. | Nothing: the same campaign tries them again. |
| "\{count} not called by this campaign" | More people were left out than one day of your cap, some are already being called by another campaign, or their calling hours are too short for a retry (a one-hour window can be). | Launch another campaign later to reach them, or widen the calling hours. |
| "Campaign paused, but \{count} batch(es) could not be confirmed stopped and may still be calling. Press Pause again." | Some batches may still be calling. | Press **Pause again**. If it repeats, contact us. |
| **Contacts and imports** | | |
| "No change (the number could not be recorded)." | Consent was not written. | Refresh and try again; contact us if it repeats. |
| "Use only letters, numbers, and underscores (and not a reserved name)." | The column name is not allowed. | Choose another name. |
| "Could not import the contacts." | The import failed. Nothing to undo. | Try again. |
| **WhatsApp** | | |
| `no_consent` (after "Not sent (Suppressed):" or in **Detail**) | No WhatsApp consent on the contact. | Record WhatsApp consent in **Contacts**. |
| `opt_out` | They opted out, or no saved contact has this exact number. | Add the contact in `+` format, or leave them opted out. |
| `frequency_cap` | 4 marketing messages already in 7 days. | Wait. |
| "Not sent (Failed): unresolved template params at index \[n]." | A variable came out empty. `index [0]` is `{{1}}`. | Fill in the contact's detail, or change the variable. |
| "We could not confirm whether this message reached WhatsApp. Check the send log before sending it again, so the person does not receive it twice." | No clear answer came back. | Check **Recent sends** before sending again. |
| "No WhatsApp account with messaging enabled + an assigned agent is available for this client." | No usable WhatsApp number. | On **Phone numbers**, switch on **Messaging enabled** and assign an agent. |
| "A valid E.164 phone (or a default\_calling\_code) is required." | The number has no country code. | Retype it with `+` and the country code. |
| "Your browser blocked the Meta window. Allow pop-ups for this site, then press Connect WhatsApp again." | A pop-up blocker stopped Meta's window. | Allow pop-ups for the portal and try again. |
| **Store connection** | | |
| "The last approval page expired before it was used. Start again when you are ready." | The 10 minutes passed. | Start again. |
| "The connection was approved on a different website than the store address entered here. Check the address and start again." | You approved it on another site. | Check **Your store address** and start again. |
| "We could not reach your store to check the connection. Check that your store is online, then start again." | Your store did not answer. | Check your store is online and start again. |
| "Too many attempts in a short time. Wait ten minutes, then try again." | More than 5 attempts in 10 minutes. | Wait ten minutes. |
| **Team and access** | | |
| "Sorry, members could not be invited. Please try again." | An address already belongs to a member. | Take that address out and send again. |
| "Invite not found or expired" | The invitation expired, was removed or used, or the email differs. | Ask an owner to renew it or invite the right email. |
| "Could not update Organization. Please try again." | Only the primary owner can save the name or logo. | Ask the primary owner. |
| "Reauthentication Required" | The portal needs a fresh sign-in. | Sign out, sign in again, and retry. |
| **Billing** | | |
| "You cannot manage billing" | You are a member. | Ask an owner. |
| "Error requesting checkout" | Payment could not start. | Try again a little later. |
| "We could not confirm your cancellation" | The cancellation may or may not have gone through. | Reload the page to check before trying again. |
## Still stuck?
Email us at [info@ruleinside.com](mailto:info@ruleinside.com). To help us answer in one reply, include:
* your **workspace name**, as shown in the portal;
* the **agent** involved, by name;
* the **date and time** of the call or message, and your time zone;
* the **caller's number, masked**, for example `+1555123****`;
* a **screenshot** of what you see, including any message;
* **what you expected, and what happened instead**.
Never send us passwords, API keys, licence keys, signing secrets or Twilio Auth Tokens. We never ask for them.
# Webhooks
Source: https://docs.ruleinside.com/webhooks
Have RuleInside send a signed message to your own system whenever a call ends, a contact opts out or consent changes.
A webhook sends a signed HTTPS request to an address on your own system (an **endpoint**) whenever something
happens: a call ends, a contact opts out, consent changes. Your system can then update your CRM, alert your team or
start its own process.
Open **Settings**, choose the **Developer** tab, and find the **Webhooks** card. The same tab holds your
[API keys](/api-keys).
Webhooks send data **out** of RuleInside. To have your other systems send contacts and events **into** RuleInside,
use an [inbound connection](/connections#inbound-connections).
## Who can do what
* **Everyone on your team** can see the endpoints and the recent deliveries.
* **Only the owner** can add or remove an endpoint, because an endpoint decides where your customer data is sent.
If your workspace is not linked to a store yet, the card asks you to connect your store first.
## Your endpoints
Each endpoint on the list shows:
* its address;
* a grey badge with its format: `normalized` for every endpoint you add on this page (an endpoint our team set up
for you may show `raw`);
* the events it receives: one badge per event, **All events**, or, for a `raw` endpoint, every completed call;
* **Added** and the date, followed by your description if you gave one;
* **Remove** (owner only).
With no endpoints you see "No endpoints yet. Add one to start receiving events."
If you see "We could not load your endpoints", the list could not be loaded at that moment. It does not mean you have
none, and an endpoint that was already receiving events is still running. Reload in a moment.
## Add an endpoint
Click **Add endpoint** (owner only) and fill in:
| Field | Rules | What it does |
| - | - | - |
| **Endpoint URL** | Required. Must start with `https://`, up to 2,048 characters, with no user name or password in it. | The address we send each message to, for example `https://example.com/webhooks`. |
| **Description (optional)** | Up to 200 characters | A note for you, shown on the list. |
| **Events** | Tick boxes, all unticked by default | The events this endpoint receives. Leave all unticked to receive every event, including ones we add later. |
Click **Add endpoint** in the dialog (available once a URL is typed). It reads "Adding…" while it works.
The address is checked before it is saved:
| Message | What to do |
| - | - |
| "The URL must not contain a username or password. Use a header for a shared secret." | Remove the user name and password from the address. |
| A refusal for an address on a private or internal network | Use a public address. Addresses that point to a private or internal network are refused. |
| "We could not look up that domain right now. Check the address, then try again." | Check the address for typing mistakes, then try again. |
| "A webhook endpoint with this URL already exists. Delete it first to re-register." | Remove the existing endpoint first, then add it again. |
| "Could not add the endpoint." | Try again; contact us if it keeps happening. |
### Copy your signing secret
When the endpoint is added, the **Your signing secret** dialog shows its signing secret. Click **Copy**, store the
secret in your system, then click **Done**.
The signing secret is shown only once and cannot be retrieved later. If you close the dialog without copying it,
remove the endpoint and add it again to get a new secret.
## The events
| Event | Sent when |
| - | - |
| `call.completed` | A call or conversation ended. It includes the channel (such as the website widget, phone or WhatsApp), the duration, the outcome, the summary, the contact, and the data your agent collected. |
| `consent.updated` | Calling or WhatsApp consent was given or withdrawn for one or more numbers. |
| `contact.opted_out` | A contact asked not to be contacted. |
| `contact.updated` | A contact was created or changed, for example through an [inbound connection](/connections#inbound-connections) or the API. |
There is no event for a finished campaign at the moment.
## What each message looks like
Each message is an HTTPS `POST` with a JSON body:
| Field | What it is |
| - | - |
| `id` | The event's id. It stays the same if the event is sent again, so use it to ignore duplicates. |
| `type` | The event, for example `call.completed`. |
| `api_version` | `v1`. |
| `created_at` | When the event was created. |
| `data` | The details of the event. |
```json Body theme={null}
{
"id": "",
"type": "contact.opted_out",
"api_version": "v1",
"created_at": "
### Check the signature
Every message carries an `X-RuleInside-Signature` header with a time (`t`) and a signature (`v0`). The signature
is an HMAC-SHA256, made with your signing secret, of the time, a dot (`.`), and the exact body of the message, as
received.
To check that a message really came from RuleInside and was not changed on the way, your system:
1. reads `t` and `v0` from the header;
2. computes the HMAC-SHA256 of `t`, a dot and the raw body, using your signing secret;
3. accepts the message only if the result matches `v0`.
Compute the signature on the body exactly as it arrived, before your system parses or reformats it.
## Deliveries and retries
* A delivery succeeds when your endpoint answers with any `2xx` code within **15 seconds**. Anything else counts as
a failure.
* A failed delivery is retried, with a longer wait each time. In practice the first retries come about 15 minutes
apart, and the longest wait between two attempts is about two hours.
* After **8** attempts the delivery is marked **dead**: it is not retried any more, and you get a notification.
A dead delivery is not sent again automatically, even after you fix your endpoint.
* Delivery is "at least once": rarely, the same event arrives twice. Ignore an `id` your system has already
processed.
* Finished deliveries (delivered or dead) are removed from the list after 30 days.
## Recent deliveries
**Recent deliveries** lists the latest 50 deliveries across all your endpoints, newest first. Each row shows the
endpoint address, the time, how many attempts were made ("N of M attempts"), an error type if there was one, and
a status:
| Status | Badge | Meaning |
| - | - | - |
| `pending` | Grey | Waiting to be sent. |
| `failed` | Red | The last attempt failed. It will be retried. |
| `delivered` | Dark | Your endpoint accepted it. |
| `dead` | Red | All attempts are used up. It will not be retried. |
The error type tells you why an attempt failed. Your endpoint's own reply is not shown, only the type:
| Error type | What your endpoint did |
| - | - |
| `auth_rejected` | Answered `401` or `403`: your system refused the request. |
| `not_found` | Answered `404`: the address does not exist. |
| `rate_limited` | Answered `408` or `429`: too busy or timed out. |
| `server_error` | Answered with a `5xx` code: an error on your server. |
| `client_error` | Answered with another `4xx` code. |
| `unreachable` | Did not answer, took too long, or the address was refused. |
With no deliveries you see "No deliveries yet." If you see "We could not load recent deliveries", the list could
not be loaded at that moment. It does not mean nothing was delivered.
## Remove an endpoint
Click **Remove** (owner only) and confirm with **Remove endpoint** ("Remove this endpoint?"). You see "Endpoint
removed.", or "That endpoint was already removed, or no longer exists."
Removing an endpoint cannot be undone. No new events are sent to it from that moment, and its signing secret stops
working and cannot be recovered. If you add the same address again later, it gets a **new** signing secret.
Messages that were still waiting to be sent to the endpoint when you removed it are not sent. Events that happen
after the removal are not sent to it either.
## Change the signing secret
There is no button to replace a signing secret in place. A new secret would take effect immediately, and every
event would fail your signature check until you updated your system. To change it:
Events that happen while the endpoint is removed are not sent to it.
Choose the same events, and copy the new signing secret.
Put the new secret into your system. Messages that were waiting when you removed the endpoint are not sent, so
events from that moment until you add it again do not reach your system.
## Common questions
Tick only `call.completed`. Leaving every event unticked means all events, including ones we add later.
Look at the error type. `auth_rejected` usually means your signature check or your own login is refusing us;
`unreachable` means your server was down or could not be reached. Fix the endpoint: new events then arrive as
normal, but dead deliveries are not sent again.
That can happen rarely. Each event has a stable `id`: ignore an `id` your system has already processed.
Remove the endpoint and add it again, then update your system with the new secret. See
[Change the signing secret](#change-the-signing-secret).
It cannot be shown again. Remove the endpoint, add it again and copy the new secret.
No. The address must be public and use `https://`. Addresses that point to a private or internal network are
refused.
# Website assistant
Source: https://docs.ruleinside.com/website-assistant
Put your agent on your website, choose how it looks and what it says, and keep it under your control.
The website assistant (the **widget**) is a voice and chat box on your website. Shoppers talk to, or type to, one
of your agents through it. Open **Website widgets** in the sidebar to create and manage your widgets.
Each widget puts **one** agent on your website, and each agent can have only one widget.
## Who can do what
* **Everyone on your team** can open the widgets list and every widget, and read all its settings.
* **Only the owner** can create a widget, change its settings, switch it on or off, rotate its embed id or delete
it. Members see the settings greyed out, without the buttons that change something, and a note that only account
owners can add or change widgets.
See [Who can do what](/roles).
The website widget is available on paid plans, and only while your account is active.
## The widgets list
The list shows one row per widget, oldest first. Each row shows:
* the **agent's name** (if your agents could not be loaded, the row shows "Agent" instead);
* a status badge: **Enabled** (the widget is switched on) or **Disabled** (it is switched off and shows nothing on
your website until you switch it back on);
* the **embed id**, in small grey letters: the identifier inside your snippet that says which widget to load (the
WordPress plugin calls it **Widget ID**);
* a pencil icon (**Edit widget**) that opens the widget editor;
* a bin icon (**Delete widget**), for the owner only.
If the page shows "Couldn't load your widgets. Please try again.", the list could not be loaded at that moment.
Your widgets are not gone; reload the page.
## Creating a widget
The **New website widget** dialog opens. Only the owner sees this button.
**Agent** lists only the agents that do not have a widget yet. The first one is selected by default.
**Allowed domain** is the website the widget may appear on, for example `https://shop.example.com`. It starts
filled with `https://`. You can add more domains later in the editor. If you leave it as `https://` or blank,
the widget is created with no allowed domain and appears nowhere until you add one.
The widget is created **switched on**, and its editor opens so you can copy the snippet and set it up.
Instead of the fields, the dialog may show:
* "You need an agent first." with a **Create one** link: create an agent first.
* "All your agents already have a widget.": every agent already has its widget. Create another agent if you want a
second, differently set up widget.
Messages you may see when creating a widget:
| Message | What to do |
| - | - |
| "The website widget is available on paid plans." | Move to a paid plan. |
| "A widget already exists for this agent." | Open that agent's existing widget instead. |
| "This agent is not fully provisioned yet. Try again shortly." | Wait a few minutes and try again. |
| "Could not create the widget." | Try again; contact us if it keeps happening. |
You also see a message if your account is not active, or if a domain is not accepted (see
[Allowed domains](#allowed-domains)).
## The widget editor
Click the pencil on a row to open the editor (page title **Website widget**). The settings are on the left and a
**Live preview** is on the right.
At the top, **Agent** shows which agent this widget uses, with a **Managed** badge if it is the store agent our
team built for you. You cannot change the agent of a widget: a widget belongs to one agent for good. To use another
agent, create a new widget for it and replace the snippet on your site.
### How saving works
Nothing you change in the editor takes effect until you click **Save changes** (bottom right). You then see
"Saved.", or a message that says what to fix (for example a colour or a domain). The only buttons that act
straight away are **Rotate embed id** and **Delete widget**, after you confirm them.
Every drop-down starts with **Default**, which means "use the widget's built-in value".
Clearing a setting does not reset it to the default today. Once a text, colour, image address or drop-down has been
saved with a value, emptying the field (or choosing **Default**) and saving again keeps the old value. To change it,
type or choose the value you want instead, for example the default wording shown on this page, or a specific
colour.
### Install
**Embed snippet** (read only) holds the two lines you paste into your site's HTML, where you want the widget to
load. It looks like this, with your own embed id:
```html Floating launcher theme={null}
```
```html Inline embed theme={null}
```
* **Copy snippet** is the copy icon at the top right of the box. It copies both lines, and the icon turns into a
tick for two seconds. Members can copy the snippet too; only the settings below it are greyed out for them.
* With **Widget mode** set to **Inline embed**, the first line also carries the inline mode and, if you filled
them in, the width and height. So changing **Widget mode**, **Width (px)** or **Height (px)** changes the snippet
itself: paste the new snippet onto your site after saving.
On WordPress with the RuleInside plugin, you do not paste the snippet. Paste only the **embed id** into the
plugin's **Widget ID** field (WordPress admin, menu **RuleInside**). See [WordPress plugin](/plugin).
#### Allowed domains
**Allowed domains** lists the websites the widget may appear on, one per line, for example
`https://shop.example.com`. The widget only shows on these websites. If the list is empty, it shows nowhere.
| Rule | Detail |
| - | - |
| Secure addresses only | Each line must start with `https://`. `http://` is accepted only for `localhost`, `127.0.0.1` and names ending in `.localhost`, for testing on your own computer. |
| The address of the site only | No page path, no `?` query, no `#` part, no user name or password. A trailing `/` is fine and is removed. |
| Tidied for you | Addresses are saved in lower case, and repeated lines are removed. |
| Up to 25 lines | More than 25 lines are refused, even if some of them repeat. |
Messages you may see: "Enter a full origin like [https://shop.example.com](https://shop.example.com).", a message that the address must use
https, or a message that it must not include a path, query or fragment.
The widget also refuses to open on its own in a browser tab: it only works inside a website that embeds it.
### Interface
| Setting | Options | Default | What it does |
| - | - | - | - |
| **Widget mode** | **Floating launcher**, **Inline embed** | Floating launcher | Floating: a launcher fixed in a corner of the screen. Inline: the full widget appears in your page exactly where you placed the snippet. Changes the snippet. |
| **Launcher size** | **Full card**, **Compact**, **Orb only** | Full card | The size of the floating launcher: a full card, a compact orb with a label, or just the round orb. Has no effect with Inline embed. |
| **Placement** | **Bottom right**, **Bottom left**, **Top right**, **Top left** | Bottom right | The corner the floating launcher sits in. |
| **Modes offered** | **Voice & text**, **Voice only**, **Text only** | Voice & text | Whether shoppers can talk, type, or both. |
| **Post-call feedback** | **After the call**, **Off** | After the call | Whether the widget asks the shopper for feedback when a conversation ends. |
| **Width (px)** | Whole number from 200 to 1200 | The widget's own size | Shown only with Inline embed. The width of the widget in your page. Changes the snippet. |
| **Height (px)** | Whole number from 200 to 1000 | The widget's own size | Shown only with Inline embed. The height of the widget in your page. Changes the snippet. |
Enter **Width (px)** and **Height (px)** as plain numbers, such as `600`. A value with letters, such as `600px`, is
ignored when you save. A number outside the range is refused with a message such as "Too big: expected number to
be \<=1200".
If you change **Widget mode** but do not paste the new snippet, your website keeps the layout of the snippet that is
already on the page.
### Appearance
| Setting | Options | Default | What it does |
| - | - | - | - |
| **Theme** | **Follow system**, **Light**, **Dark** | Light | Follow system matches the shopper's light or dark device setting. The colours below override the theme. |
| **Orb color** | A colour like `#1a2b3c` | Widget default | The animated round orb. |
| **Background** | A colour like `#1a2b3c` | Widget default | The background of the widget panel. |
| **Text** | A colour like `#1a2b3c` | Widget default | The text colour. |
| **Border** | A colour like `#1a2b3c` | Widget default | The border colour. |
| **Button** | A colour like `#1a2b3c` | Widget default | The background of the buttons. |
| **Button text** | A colour like `#1a2b3c` | Widget default | The text on the buttons. |
| **Corner radius** | **None**, **Small**, **Medium**, **Large**, **Extra large** | Widget default | How rounded the corners are. |
| **Shadow** | **None**, **Small**, **Medium**, **Large** | Widget default | The shadow around the widget. |
| **Font** | **Geist**, **Inter**, **Lora**, **Monospace** | Widget default | The typeface. |
| **Avatar image URL** | An address starting with `https://`, up to 600 characters | The orb | An image shown as the launcher icon instead of the orb. |
Each colour must be a 6-digit hex colour, such as `#1a2b3c`, or you see "Use a 6-digit hex color like #1a2b3c."
You can type it or pick it from the swatch.
If **Button** and **Button text** are too close to read easily, you see "Low contrast between the button and its
text". It is a warning only: you can still save, but shoppers may find the buttons hard to read.
### Content
| Setting | Limit | Default wording | What it is |
| - | - | - | - |
| **Assistant name** | 60 characters | "AI Assistant" | The name at the top of the widget. |
| **Launcher prompt** | 60 characters | "Need help?" | The short line on the launcher. |
| **Start-call button** | 40 characters | "Start a call" | The button that starts a voice call. |
| **Returning-shopper greeting** | 200 characters | "Ciao \{name}! You are speaking with an AI assistant and this call may be recorded. How can I help?" | What the agent says first to a signed-in shopper it recognises. |
| **Show a terms & conditions gate** | On or off | On | Before their first conversation, the shopper must accept a consent notice. |
| **Terms text** | 2,000 characters | A notice that conversations with the AI assistant may be recorded, stored and processed by third-party providers as described in the Privacy Policy | The consent notice. Shown only while the gate is on. |
#### The returning-shopper greeting
This greeting welcomes a signed-in shopper your store recognises, by name. Write `{name}` where their first name
should go. It is used only when a recognised shopper starts a **voice** conversation, never for guests and never
for text chats.
The greeting must tell the shopper **both** that they are speaking with an AI and that the call may be recorded.
To count, it needs:
* an AI mention: "AI", "A.I.", "artificial intelligence", "virtual assistant" or "automated assistant" (Italian
"IA", "assistente virtuale" and "intelligenza artificiale" also count);
* a recording mention: any word containing "record" (or, in Italian, "registrat").
If the greeting does not mention both, you can still save it, but the widget does not use it: your agent's own
first message is used instead, and the portal does not warn you. Curly braces other than `{name}` are removed
before the greeting is spoken.
### Status
* **Widget enabled**: switch it off and click **Save changes** to stop the widget showing on every website at once.
The embed id stays the same, so switching it back on brings the widget back with no change to your website.
* **Rotate embed id** (owner only): creates a new embed id. You are asked to confirm ("Rotate the embed id?").
Afterwards a message reminds you to update the snippet on your site.
* **Delete widget** (owner only): deletes the widget after you confirm, then returns you to the list.
**Rotate embed id** cannot be undone. The old embed id stops working at once, so every website still using the
old snippet, and every WordPress plugin whose **Widget ID** still holds the old id, shows nothing until you update
it. Use it if your snippet was copied onto a website where it should not be. It needs an active account.
## Deleting a widget
Click the bin icon on the list, or **Delete widget** in the editor, and confirm ("Delete this widget?"). You see
"Widget deleted."
Deleting a widget cannot be undone. It stops loading on your site straight away. Your agent is not affected, and
you can create a new widget for the same agent, but it gets a **new** embed id: paste the new snippet (or the new
**Widget ID** in the plugin) on your site.
Deleting an **agent** also deletes its widget.
## The live preview
The right side of the editor shows **Live preview**: the real widget, updated a moment after each change, without
interrupting a conversation in progress. **Width (px)** and **Height (px)** are not shown in the preview.
The preview is a real conversation with your agent. Your browser asks for microphone access, and messages you type
are really sent. Like a conversation on your website, it needs the widget switched on, your account active and on a
paid plan, and it counts towards the daily conversation limit.
## Limits that protect your widget
* **30 conversation starts a minute for each widget.** Above that, new conversations cannot start until the minute
is over.
* **A daily number of conversations for your account**, shared by all your widgets and set from your plan. The
day resets at midnight UTC, not at your local midnight. Your portal notifies you when half of the day's
conversations are used and when all of them are; new conversations then start again the next day. See
[Plans, limits and what happens at the limit](/limits).
* If you see "Your website assistant is paused. Contact support to turn it back on.", contact us.
Every new conversation checks again that the widget is switched on and that your account is active and on a paid
plan, so a paused account's widget stops at once.
## Common questions
Check these in order:
1. Is the exact address of your site, starting with `https://`, listed in **Allowed domains**?
2. Is **Widget enabled** switched on, and did you click **Save changes**?
3. Was the embed id rotated after you pasted the snippet? Then paste the new one.
4. Is your account active and on a paid plan?
5. Did you reach today's conversation limit? Look for a notification in your portal.
6. Are you opening the widget's own address in a browser tab? That is refused on purpose: open your website
instead.
Yes. Add each website on its own line in **Allowed domains** (up to 25), save, and paste the same snippet on
each site.
No, each agent has one widget. If you want a second widget set up differently, create another agent for it.
No. Create a widget for the other agent and replace the snippet on your site.
Make sure you clicked **Save changes**. If you changed **Widget mode**, **Width (px)** or **Height (px)**, the
snippet itself changed: copy it again and paste it onto your site.
Clearing a field does not reset it today. Type the wording or value you want instead, for example the default
wording listed on this page.
It must say both that the shopper is speaking with an AI and that the call may be recorded, using the words
listed under [The returning-shopper greeting](#the-returning-shopper-greeting). Otherwise your agent's own first
message is used. It is also used only for signed-in shoppers your store recognises, and only for voice.
Switch **Widget enabled** off and click **Save changes**. Switch it back on later: the snippet on your site does
not need to change.
Only the owner of the account can change widgets. Ask an owner of your team.
# WhatsApp
Source: https://docs.ruleinside.com/whatsapp
Connect your WhatsApp number, register your templates, send messages and follow what happened to them.
RuleInside can message your customers on WhatsApp: automatic follow-ups after a chat, follow-ups from your
automations, WhatsApp campaigns, and single messages you send yourself. Setting it up takes three steps.
On **Phone numbers**, connect the WhatsApp Business number your customers message you on. See
[Connecting your number](#connecting-your-number) below.
Create and approve your templates in Meta's WhatsApp Manager, then register them under
[WhatsApp templates](/whatsapp-templates).
Send one message from [WhatsApp send](/whatsapp-send), or let your follow-ups, automations and campaigns use the
templates you registered.
WhatsApp only lets a business start a conversation with a template Meta has approved. RuleInside cannot send
free-form text first, and it does not create templates: you create them in Meta and register them in RuleInside.
## Who can do what
Any member of your team can look at the WhatsApp pages. Only the **account owner** can connect a number, change
WhatsApp number settings, register, edit or delete templates, and send messages. A member sees a note saying an owner
is needed, and the buttons they cannot use are greyed out or hidden.
## Connecting your number
Open **Phone numbers**. In the **WhatsApp** panel, the owner sees a **Connect your WhatsApp number** card when:
* your workspace is linked to a store;
* no WhatsApp number is connected yet, or your connection to Meta needs to be repaired.
The card is not shown while a number is connected and working, because connecting again would replace a working
connection. If self-connection is not available for your account, the card says so instead: contact support and our
team connects your number for you.
Before the button, the card lists what changes when you connect:
* You keep using the WhatsApp Business app on your phone, and it has to be opened at least once every 13 days. If
you uninstall it, the connection drops.
* Some things change on your phone for good: disappearing messages and view-once are turned off in one-to-one chats,
live location sharing stops, broadcast lists become read-only, group chats are not carried over, and your profile
photo can no longer be changed.
* Keep the app open for a few minutes after you connect while your existing conversations sync.
After connecting, **add a payment method to your WhatsApp Business Account in Meta**. Meta bills you directly for the
messages you start, and blocks them until a payment method is added.
A Meta sign-in window opens. The button reads **Connecting…** while it is open. If your browser blocks the
window, allow pop-ups for the portal and click **Connect WhatsApp** again.
Sign in with the Meta account that owns your WhatsApp Business number, and pick the business account and number
in Meta's screens. RuleInside never sees your Meta password.
When Meta finishes, the window shows **Finishing up. This window will close on its own.** and closes. Do not
reload the portal while the Meta window is open, or the attempt is lost and you have to start again.
You see **WhatsApp connected.** and your number appears in the panel with the badge **Connected through Meta**.
While the Meta window is open, a **Start over** button appears next to **Connect WhatsApp**. Use it if the window got
lost. If you start over and then finish in the old window anyway, that connection is not saved: click **Connect
WhatsApp** and do it again.
### If you manage more than one WhatsApp Business Account
If your Meta login manages several WhatsApp Business Accounts, the card asks **Which WhatsApp number is this store’s?**
and lists each one: the account name (or **Unnamed WhatsApp account**), then the number and the verified business
name.
* Pick the number your customers message this store on, then click **Connect this number**. Nothing is connected
until you do.
* **Cancel** closes the choice without connecting anything.
The choice is kept for 15 minutes. After that, connect again from the start.
### Messages you may see while connecting
| Message | What to do |
| - | - |
| **You closed the Meta window before finishing. Nothing was changed.** | Click **Connect WhatsApp** again when you are ready. |
| **Your browser blocked the Meta window. Allow pop-ups for this site, then press Connect WhatsApp again.** | Allow pop-ups for the portal, then try again. |
| **Meta didn't finish the sign-in, so nothing was connected. You can try again.** | Try again. |
| That WhatsApp Business Account is already connected to another client on this platform. | Contact support to have it released first. |
| That WhatsApp Business Account has no phone number yet. | Add the number during Meta's sign-up and connect again. |
| That phone number is not on that WhatsApp Business Account. | Check you picked the right account and number, and connect again. |
| **Could not complete the WhatsApp connection.** | Try again; contact support if it keeps happening. |
If you open the page that finishes the connection on its own (from a bookmark, for example), it says there is
nothing to finish there: go back to **Phone numbers** and click **Connect WhatsApp**.
## Your number on Phone numbers
### A number connected through Meta
The panel shows the number with the **Connected through Meta** badge. There is nothing to change here: its settings
are managed by RuleInside. To disconnect, use **Settings > Account > Business Platform** in your WhatsApp Business
app.
A status line shows what WhatsApp reports about the number:
| Status line | Meaning |
| - | - |
| **WhatsApp lets this number send messages.** | All is well. |
| **This number can send messages, with a limit:** | Sending works, with the limit Meta describes underneath. |
| **WhatsApp is blocking messages from this number. Until it is fixed, the messages you send wait here and are not sent:** | Meta's reasons and fixes are listed underneath (often a missing payment method). Messages wait as **Queued** and go out once Meta allows it. A message still waiting 7 days after it was queued is marked **Failed** instead. |
| **We couldn't check this number's status with WhatsApp right now.** | Refresh later. |
If your connection to Meta was left incomplete, a red message explains it cannot be used yet. The owner can reconnect
with the **Connect your WhatsApp number** card below it; a member is asked to contact a workspace owner.
### A number our team connected for you
A number without the **Connected through Meta** badge shows its number and business name, and the owner can change:
| Setting | What it does | Who can change it |
| - | - | - |
| **Agent** | The agent assigned to this WhatsApp number. | Owner |
| **Messaging enabled** | Turns WhatsApp messaging on or off for the number. | Owner |
Click **Save** to apply your changes. To send any WhatsApp message, the number needs **Messaging enabled** switched on
and an agent assigned. A **Token expired** badge means the number needs reconnecting: contact support.
## Templates
You create, word and submit templates in Meta's WhatsApp Manager. Once Meta approves one, you register it under
**WhatsApp templates**: its purpose (immediate follow-up, event follow-up, re-engagement for campaigns, or manual
send), its exact name, language code and Meta category, the number it sends from, and what fills each numbered
blank. Each purpose holds one message template; saving another for the same purpose replaces it, and changing a
template's purpose or kind saves a copy and leaves the original. The **At WhatsApp** column shows each template's
live status at Meta.
Read [WhatsApp templates](/whatsapp-templates) for every field, status and button.
## Sending and the send log
**WhatsApp send** sends one template message to one number, after a confirmation. The person must be a saved contact
with WhatsApp consent recorded: a number that is not a saved contact is refused. **Recent sends** shows your newest
100 WhatsApp messages from every source, with their status in RuleInside and, for a number connected through Meta,
what WhatsApp reported (delivered, read, or not delivered with WhatsApp's reason).
Read [Sending a WhatsApp message](/whatsapp-send) for every field, status and reason.
## The rules every message follows
Every WhatsApp message, whether a follow-up, campaign or manual send, is checked first. A person who fails any check
is not messaged, and if a check cannot be completed at that moment, the message is not sent.
* Your store has completed its confirmation under **Settings › Calling consent**.
* The person has not opted out of WhatsApp, and is not on your WhatsApp do-not-contact list.
* The person is a saved contact with WhatsApp consent recorded. Consent to calls does not count.
* **Weekly marketing limit**: at most 4 Marketing-category WhatsApp messages per contact in any rolling 7 days.
Utility and Authentication templates are not counted.
* **Marketing quiet hours**: Marketing templates are sent only during the person's calling hours, and wait until then
otherwise. An immediate follow-up (in a chat the shopper started) does not wait when it goes within 24 hours of
that chat; after that, it waits like any other Marketing message. Utility or Authentication templates do not wait.
A Marketing message waiting only because the contact has no time zone is not sent once it has been queued for
7 days.
* **Marketing stopped**: if WhatsApp refuses a Marketing message because the person told WhatsApp to stop marketing
messages from your business, that number is sent no more Marketing templates from your store. Utility and
Authentication templates still go. If the person later turns your marketing messages back on in WhatsApp, the stop
is lifted. Both come from what Meta reports, so they apply to a number connected through Meta.
* Opt-out, do-not-contact and consent are checked again just before each message goes out.
See [Why a send is suppressed](/whatsapp-send#why-a-send-is-suppressed) for the reasons shown in the send log, and
[How calling rules protect your customers](/outbound-rules) for the rules across calls and messages.
## STOP and START
These replies are read from the messages Meta passes to RuleInside, so they apply to a number connected through Meta.
### When a customer replies STOP
A customer who replies with a message whose whole text is **STOP** is opted out of WhatsApp messages and calls from
your store.
* Capital letters, punctuation, emoji and formatting around the word do not matter: "STOP", "stop", "Stop." and
"STOP 🛑" all count.
* Tapping a quick-reply button on your template whose label is STOP also counts.
* A longer message does not: "stop calling me", "please stop" or "STOP STOP" are not treated as an opt-out.
* Only the customer's own replies count. Typing STOP from your own WhatsApp Business app opts nobody out.
The opt-out covers **WhatsApp and calls**: the number is opted out on the contact and added to your do-not-contact
list for both. Email is not affected. A message or call already waiting for that number is stopped when its turn
comes.
### When a customer replies START
A customer who opted out can opt back in by replying with a text message whose whole text is **START** (the same
rules as STOP, but a button tap does not count). Their WhatsApp opt-out is lifted and their WhatsApp consent is
recorded again. Calls stay stopped: START brings back WhatsApp messages only.
Nothing changes when:
* the number never opted out (START brings back someone who left; it does not sign up someone new);
* something newer than the START says not to contact them, such as a later STOP, a button reply, or a newer opt-out;
* the START could not be processed within 48 hours of arriving.
### The confirmation reply
When a customer's STOP is recorded, or their START brings them back, RuleInside replies once in that chat with a
short message:
| After | The reply (in English) |
| - | - |
| **STOP** | "You will not receive calls or WhatsApp messages from us any more. Reply START to receive WhatsApp messages again." |
| **START** | "You will receive our WhatsApp messages again." |
* It is sent from your number connected through Meta, and only when the customer wrote to that number.
* It is sent only within 24 hours of the customer's message. Later, no reply is sent.
* A START that changes nothing (see the list above) gets no reply.
* It is written in your store's language: the language of your WhatsApp templates when they all use one language,
otherwise the main language of your store's country. English, Italian, Spanish, French, German and Portuguese are
available; anything else gets English. The word START stays START in every language.
* If the reply cannot be sent, the customer's STOP or START still stands.
**You cannot reverse an opt-out.** Only the customer can, by replying START. Neither you nor our support team can
lift it from the portal.
## Common questions
No. Create and submit them in Meta's WhatsApp Manager. Once Meta approves one, register it under **WhatsApp
templates** with its exact name and language. See [WhatsApp templates](/whatsapp-templates).
The most common reasons: the number is not a saved contact, the contact has no WhatsApp consent recorded, they
opted out, or they already had 4 marketing messages this week. The send log shows the reason in **Detail**. See
[Why a send is suppressed](/whatsapp-send#why-a-send-is-suppressed).
No. Only the customer can opt back in, by replying START.
Yes, one short confirmation in your store's language, when your number is connected through Meta and the reply
can go within 24 hours of their message. See [The confirmation reply](#the-confirmation-reply).
Yes. A STOP reply opts them out of WhatsApp messages and calls. Email is not affected. Replying START later brings
back WhatsApp messages only; calls stay stopped.
It shows only to the account owner, only when no working number is connected, and only when self-connection is
available for your account. If a number is already connected and working, the button is hidden on purpose.
Check the **Detail** column. A Marketing template waits for the person's calling hours. If your number's status
on **Phone numbers** says WhatsApp is blocking messages, fix it in Meta (often by adding a payment method) and
the messages are sent.
When your number is connected through Meta, Meta bills you directly for the messages you start, and blocks them
until you add a payment method to your WhatsApp Business Account in Meta.
In your WhatsApp Business app, go to **Settings > Account > Business Platform**.
No. Members can see the WhatsApp pages and the send log, but only the account owner can connect numbers, manage
templates and send.
# Sending a WhatsApp message
Source: https://docs.ruleinside.com/whatsapp-send
Send one template message to one customer, follow it in the send log, and understand why a send was held or suppressed.
**WhatsApp send** sends a single WhatsApp template message to one number, to test a template or to reach one
customer, and shows your recent sends. It sends one message to one number each time. There is no free-text message
and no bulk sending on this page: to reach many people, use [Campaigns](/campaigns).
## Before you send
Check these four things first. A send that fails any of them is not delivered.
Register the template under [WhatsApp templates](/whatsapp-templates), with **Kind** set to **Message** and
**Enabled** on.
The person must be a saved contact in [Contacts](/contacts), with their number in international format starting
with `+` (for example `+15551234567`).
The contact must have WhatsApp consent recorded. Consent to calls does not count for WhatsApp.
An owner must have completed the confirmation under **Settings › Calling consent**. Until then, no calls or
messages are sent.
**A number that is not a saved contact is refused.** RuleInside looks up the contact by the exact number you type.
If there is no contact with that number, the message is not sent and the send log shows **Suppressed** with the
reason `opt_out`. Add the contact first, then record their WhatsApp consent, then send.
## Sending a message
Type the recipient's number in **To (phone number)**, in international format with the country code.
Pick it under **Template**.
If the template uses send-time values, type each one under **Template variables**.
This does not send yet: it opens a confirmation.
The message is sent, or the page tells you why it was not.
## Every field
Only the account owner can fill in and send this form.
| Field | What it does | Allowed values and limits | Default | Who can change it |
| - | - | - | - | - |
| **To (phone number)** | The recipient's WhatsApp number. | International format, including the country code: it must start with `+` or `00`. Spaces, dashes and brackets are fine. Letters (for example an extension such as `x12`) are not allowed, and a number without `+` or `00` is refused, because no default country is assumed. The example in the box is `+15551234567`. | Empty | Owner |
| **Template** | Which registered template to send. Each option shows the purpose followed by the template name. | Only templates whose **Kind** is **Message** and which are **Enabled**. | The first template in the list | Owner |
| **Template variables** | One box for each send-time value the chosen template uses, labelled with that value's name. If one name fills two blanks, there is one box. Shown only when the template has send-time values. | Every box must be filled. | Empty. The boxes are cleared whenever you change the template. | Owner |
Values set as **Fixed text** on the template are not shown here: they are the same on every send. Values set as
**Contact field** are not shown either: they are filled from the contact whose saved number matches the number you
typed. When a template uses contact fields, the form says so: "This template uses contact fields (e.g. name). Those
are filled from the contact matching this number - if no matching contact exists, the send is blocked."
Contact details are matched on the number exactly as saved. Save your contacts' numbers in international format
starting with `+`, so they match what you type here.
If the number cannot be used, a red message says: "A valid E.164 phone (or a default\_calling\_code) is required."
Retype the number starting with `+` and the country code.
### What the page shows instead of the form
* **You have no enabled WhatsApp message template yet.** Register a message template, or enable an existing one,
under **WhatsApp templates**. Call permission templates cannot be used to send a message.
* **We couldn't load your WhatsApp templates right now.** Refresh the page; contact support if it keeps happening.
* **This workspace isn't linked to a store yet.** No form and no log are shown. Contact support.
* A member who is not the owner sees a note that only account owners can send WhatsApp messages. The **Recent
sends** log is still visible to them.
## Buttons and the confirmation
| Button | What happens |
| - | - |
| **Send message** | Greyed out until a number is typed, a template is chosen and every variable box is filled. It does not send: it opens the confirmation. |
| **Cancel** (in the confirmation) | Closes the confirmation. Nothing is sent. |
| **Send now** (in the confirmation) | Sends the message. It reads **Sending…** while it works, and the confirmation cannot be closed until it finishes. |
The confirmation is titled **Send this WhatsApp message?** and reminds you that this sends a **real WhatsApp
message** to the number you typed, may incur a messaging charge, and is skipped if the number is opted out, on a
do-not-contact list, or has no WhatsApp consent.
**A sent message cannot be recalled.** Every click of **Send now** sends a new message: a second click is not
recognised as a repeat. After each completed send the form clears the number and the variables, so you do not send
the same message twice by accident.
## What happens after Send now
RuleInside checks your account is active, finds the enabled template, finds the contact by number, runs the checks
described in [Why a send is suppressed](#why-a-send-is-suppressed), fills in the variables, picks the sending
number, and tries to send straight away. The checks run again just before the message goes out.
A message appears at the bottom of the screen with the result:
| What happened | Message you see |
| - | - |
| WhatsApp accepted the message | **Message Sent - it's on its way.** |
| Waiting in the queue to be sent | **Message Queued - it's on its way.** |
| Being sent right now | **Message Sending - it's on its way.** |
| Held until later (calling hours, missing time zone, or WhatsApp blocking your account) | **Queued, not sent yet (…)**, with the reason in brackets, for example "held: outside the calling hours where this person is; it is sent when they open" |
| Refused by the checks | **Not sent (Suppressed): …**, with a reason such as `no_consent`, `opt_out`, `dnc` or `frequency_cap`. See [the reasons](#reasons-you-may-see). |
| A variable came out empty | **Not sent (Failed): unresolved template params at index \[n].** |
| WhatsApp refused the message | **Not sent (Failed): Failed.** |
| WhatsApp asked us to slow down; it will be retried | **Not sent (Retrying): Retrying.** |
| No clear answer came back | **We could not confirm whether this message reached WhatsApp. Check the send log before sending it again, so the person does not receive it twice.** |
| The request could not be made | A red message with the reason, or **Could not send the WhatsApp message.** |
Other refusals you may see in red:
* **This account is not active; outbound WhatsApp is disabled.** Your account is suspended, on hold or not yet set
up. Contact support.
* **No WhatsApp account with messaging enabled + an assigned agent is available for this client.** On **Phone
numbers**, check that your WhatsApp number has **Messaging enabled** switched on and an agent assigned.
* **No enabled WhatsApp message template configured for purpose …** The template you chose is no longer enabled.
Enable it under **WhatsApp templates**.
## Recent sends
Under **Recent sends** you see your newest 100 WhatsApp sends, newest first. The log covers every WhatsApp message
your store sends, not only the ones sent from this page: automatic follow-ups, event follow-ups, campaigns and manual
sends. Every member can see it.
If nothing has been sent yet it says **No WhatsApp sends yet.** If the log cannot be loaded, refresh the page and
contact support if it keeps happening.
| Column | What it shows |
| - | - |
| **When** | When the message was created, in your own browser's time zone. |
| **To** | The recipient's number. |
| **Origin** | What sent it: **Immediate follow-up**, **Event follow-up**, **Re-engagement**, **Manual** or **Campaign**. Follow-ups after campaign calls show their own names, for example "Call Answered Followup" and "Call Missed Followup". |
| **Status** | Where the message is in RuleInside. See below. |
| **Delivery** | What WhatsApp reported after accepting it. See below. |
| **Detail** | One line, cut short; hover to read it all. It shows WhatsApp's failure reason and code, the reason a send was suppressed, a hold or error message, or WhatsApp's reference for the message. A dash when there is nothing to show. |
### Status
| Status | Colour | Meaning | What happens next |
| - | - | - | - |
| **Queued** | Grey | Waiting to be sent: due for its first try, held for the person's calling hours, or held while WhatsApp blocks your account. **Detail** shows the reason when it is held. | Sent when it is due. |
| **Sending** | Grey | Being sent right now. | Becomes **Sent**, **Retrying**, **Failed** or **Unconfirmed**. |
| **Sent** | Green | WhatsApp accepted the message. This is not proof it reached the phone: see **Delivery**. | Nothing more. |
| **Retrying** | Grey | An attempt failed in a way that is safe to repeat (for example WhatsApp asked us to slow down), or a check could not be read at that moment. | Tried again after about 5, 10, 15 and then 30 minutes. After 5 attempts it becomes **Failed**. |
| **Suppressed** | Red | Refused by the checks: the person may not be contacted (or the campaign was cancelled). **Detail** shows the reason. | Never sent. Sending again creates a new attempt that is checked afresh. |
| **Failed** | Red | Will not be sent: WhatsApp refused it, a variable was empty, the retries ran out, or it was queued 7 days or more ago and was still waiting for WhatsApp to unblock your account or for the contact's time zone. | Nothing more. Our team is notified of most of these. |
| **Unconfirmed** | Red | The message may have reached WhatsApp, but no clear answer came back. | Not retried automatically, so the person is not messaged twice. Our team is notified. Check before sending again. |
### Delivery
Delivery reports come from Meta, so they appear only when your WhatsApp number is connected through Meta (it shows
the **Connected through Meta** badge on **Phone numbers**). For any other number, **Delivery** always shows **No
report from WhatsApp**.
| Delivery | Colour | Meaning |
| - | - | - |
| **Delivered** | Green | WhatsApp reports it reached at least one of the person's devices. |
| **Read** | Green | WhatsApp reports the person read it. |
| **Accepted by WhatsApp** | Grey | WhatsApp accepted it but has not yet reported it delivered. |
| **Not delivered** | Red | WhatsApp reports it could not be delivered. **Detail** shows WhatsApp's reason and code. It is not sent again automatically. |
| **No report from WhatsApp** | Grey | Sent, or possibly sent, but no report has arrived. |
| **Not sent** | Grey | The message never went to WhatsApp (its status is **Queued**, **Retrying**, **Suppressed** or **Failed**). |
A later report never moves a message backwards: for example, **Delivered** is not replaced by **Accepted by
WhatsApp**, and **Read** replaces everything else.
## Why a send is suppressed
These checks apply to every WhatsApp message your store sends: manual sends, follow-ups and campaigns. If a check
cannot be completed at that moment, the message is **not** sent.
1. **Your account is active.** A suspended account, or one not yet set up, sends nothing.
2. **Your store has confirmed its calling consent** under **Settings › Calling consent**, for the country it trades
in.
3. **The person has not opted out of WhatsApp.** A number with no saved contact counts as opted out.
4. **The number is not on your WhatsApp do-not-contact list.**
5. **The contact has WhatsApp consent recorded.** Consent to calls is separate and does not count.
6. **The weekly marketing limit.** For a template whose **Meta category** is Marketing, a contact receives at most
**4** marketing WhatsApp messages in any rolling **7 days**, across campaigns, follow-ups and manual sends.
Messages that are queued, sending, retrying, sent or unconfirmed count; suppressed and failed ones do not. Utility
and Authentication templates are not counted or limited.
7. **Marketing quiet hours.** A Marketing template is sent only during the calling hours where the person is (your
store's calling hours, or a campaign's own). Outside them the message waits as **Queued** until the hours open.
If the contact has no time zone, or the calling hours are not valid, it also waits and says so. A message that
waits only because the contact has no time zone, and was queued 7 or more days ago, is not sent: its status
becomes **Failed** and **Detail** says why. An immediate follow-up (a reply in a chat the shopper started) does
not wait when it goes within 24 hours of that chat; after that, it waits like any other Marketing message. Utility
and Authentication templates do not wait.
8. **Marketing stopped.** If WhatsApp refused an earlier Marketing message because the person told WhatsApp to stop
marketing messages from your business, that number gets no more Marketing templates from your store: the send is
**Suppressed** with the reason `marketing_stopped`. This is checked when the message is queued and again just
before it goes out. Utility and Authentication templates still go. If the person turns your marketing messages
back on in WhatsApp, Meta tells RuleInside and the stop is lifted. A stop recorded after they turned them back on
stays. This applies to a number connected through Meta.
9. **Checked again at the last moment.** Opt-out, do-not-contact and consent are checked again just before the
message goes out, so a STOP that arrives while a message is waiting stops it.
Two more things can hold a message:
* **WhatsApp is blocking your account.** For a number connected through Meta, if Meta blocks your account from
sending (for example because no payment method was added), messages wait as **Queued** with a detail beginning
"held: Meta is blocking this WhatsApp account from sending". RuleInside checks again every 15 minutes and sends
them once Meta allows it. A message that has waited 7 days or more is marked **Failed** instead of being sent late.
Fix the account in Meta's WhatsApp Manager; the **Phone numbers** page shows what Meta reports.
* **Pace.** RuleInside sends at most 200 WhatsApp messages per store per hour.
### Reasons you may see
| Reason in Detail or in the message | What it means | What to do |
| - | - | - |
| `no_consent` | The contact has no WhatsApp consent recorded. | Record WhatsApp consent on the contact in [Contacts](/contacts). |
| `opt_out` | The person opted out of WhatsApp, **or** there is no saved contact with this exact number. | If they opted out, only they can opt back in, by replying START. If the contact is missing, add it with the number in `+` format and record WhatsApp consent. |
| `dnc` | The number is on your WhatsApp do-not-contact list. | If it was added by a STOP reply, only the customer can reverse it, by replying START. See [STOP and START](/whatsapp#stop-and-start). |
| `frequency_cap` | A Marketing template, and the contact already had 4 marketing WhatsApp messages in the last 7 days. | Wait until the oldest of those is more than 7 days old, or use a template that Meta itself categorises as Utility, when the message really is one. |
| `marketing_stopped` | A Marketing template, and the person told WhatsApp to stop marketing messages from your business. | Nothing. Only the person can turn your marketing messages back on, in WhatsApp. Utility and Authentication templates still reach them. |
| `no_store_attestation` | Your store has not completed its calling consent confirmation. | An owner completes **Settings › Calling consent**. |
| `error` | A check could not be read at that moment. Nothing was sent. | Try again later. |
| `do_not_call` | On a follow-up after a call, the person asked not to be called. | Nothing. |
| `held: …` | The message is waiting (calling hours, missing time zone, or WhatsApp blocking your account). Its status is still **Queued**. | Follow the text: set the contact's time zone, fix the calling hours, or fix the account in WhatsApp Manager. |
| `not sent: queued 7 or more days ago, and this contact still has no time zone; set the time zone and send it again` | A Marketing message waited 7 days for the contact's time zone and was not sent. Its status is **Failed**. | Set the contact's time zone, then send the message again. |
| `unresolved template params at index [n]` | A variable came out empty. The number counts from 0, so `index [0]` is `{{1}}`. | Fill in that detail on the contact, or change the variable's source on the template. |
## STOP and START
A customer who replies **STOP** to your WhatsApp messages is opted out of WhatsApp messages and calls from your store,
and any message waiting for them is suppressed when its turn comes. You cannot reverse an opt-out from the portal: only the
customer can, by replying **START**. See [STOP and START](/whatsapp#stop-and-start) for the details.
## Common questions
RuleInside needs a saved contact with this exact number and WhatsApp consent recorded on it. Being a customer of
your store is not enough, and consent to calls does not cover WhatsApp. Record WhatsApp consent on the contact,
then send again.
There is probably no saved contact with this exact number. A number with no contact counts as opted out. Add the
contact in international format (for example `+15551234567`), record WhatsApp consent, and send again.
The template is in the Marketing category, so it waits for the person's calling hours and is sent then. You do
not need to do anything. If the reason mentions a missing time zone, set the contact's time zone: a message still
waiting for one 7 days after it was queued is not sent.
**Sent** means WhatsApp accepted the message. Look at **Delivery**: **Not delivered** shows WhatsApp's reason in
**Detail**, and **No report from WhatsApp** means no report arrived (always the case when your number is not
connected through Meta). A common reason is WhatsApp blocking your account, for example because no payment
method was added: check your number's status on **Phone numbers**.
Not straight away. The message may have arrived. RuleInside does not retry it on its own, so the person is not
messaged twice, and our team has been notified. Check with the customer, or contact support, before sending
again.
A variable came out empty, usually because the contact is missing the detail the template uses (such as a name).
Fill in that detail on the contact, or change the variable on the template to fixed text or a send-time value.
At most 4 Marketing-category WhatsApp messages per contact in any rolling 7 days, across campaigns, follow-ups
and manual sends. Utility and Authentication templates are not counted.
It stays grey until you type a number, choose a template, and fill in every box under **Template variables**.
Only templates whose kind is **Message** and which are **Enabled** can be sent. Check the template under
**WhatsApp templates**.
Not from this page: it sends one message to one number. Use [Campaigns](/campaigns) to reach many people.
The confirmation warns that a send may incur a messaging charge. When your number is connected through Meta,
Meta bills you directly for the messages you start.
No. Members can read **Recent sends**, but only the account owner can send.
No. WhatsApp requires a pre-approved template to start a conversation, so RuleInside cannot send free-form text.
# WhatsApp templates
Source: https://docs.ruleinside.com/whatsapp-templates
Register the templates Meta has approved, tell RuleInside what each one is for, and read their status at Meta.
WhatsApp only lets a business start a conversation with a template that Meta has approved. Free text cannot be sent
first. **WhatsApp templates** is where you tell RuleInside which of your approved templates to use, and for which job.
**You create and approve templates in Meta, not in RuleInside.** Write the template (its header, body, buttons and
footer) in Meta's WhatsApp Manager and submit it there. Once Meta approves it, register it here by typing its exact
name and language and saying what fills each numbered blank. There is no field for the wording anywhere in the
portal: to change the wording, edit the template in WhatsApp Manager.
## Before you start
In Meta's WhatsApp Manager, write the template and submit it for review.
Meta reviews each template, and each language of it, separately.
The template's exact name, its language code (for example `en_US` or `it`), its category (Marketing, Utility or
Authentication), and how many numbered blanks (`{{1}}`, `{{2}}` and so on) its body has.
Open **WhatsApp templates**, click **New template** and fill in the form described below.
## The templates list
The page lists every template you have registered, ordered by purpose and then kind.
| Column | What it shows |
| - | - |
| **Purpose** | The job the template does: **Immediate follow-up**, **Event follow-up**, **Re-engagement (campaigns)** or **Manual send**. |
| **Kind** | **Message** or **Call permission**. |
| **Template name** | The name exactly as you registered it. |
| **Language** | The language code exactly as you registered it, for example `en` or `en_US`. |
| **Category** | The Meta category you chose: **Marketing**, **Utility** or **Authentication**. Utility and Authentication are highlighted in amber, as a reminder that they are not held to the weekly marketing limit. |
| **At WhatsApp** | The template's status at Meta, read live each time the page opens. See [Status at WhatsApp](#status-at-whatsapp). |
| **Status** | **Enabled** (green) or **Disabled** (grey): whether RuleInside uses the template. |
| (no heading) | The **Edit** and **Delete** buttons for that template. |
Instead of the list, you may see:
* **This workspace isn't linked to a store yet.** Your workspace has no store attached. The **New template** button
is shown but cannot be used. Contact support.
* **No templates yet. Add one to enable WhatsApp follow-ups and campaigns.** Nothing is registered yet.
* A red message saying the templates could not be loaded. Nothing was changed: refresh the page, and contact support
if it keeps happening.
## Registering a template
Click **New template**. A panel titled **New WhatsApp template** opens on the right. Nothing is saved until you
click **Save template**.
Pick what the template is for under **Purpose**, and leave **Kind** on **Message** unless you are registering a
call permission template.
Type the **Template name** and **Language code** exactly as WhatsApp Manager shows them.
Pick the same **Meta category** that Meta gave the template.
Under **Template variables**, click **Add variable** once for each numbered blank in the template body, in order,
and choose what fills it.
Leave **Enabled** on and click **Save template**. You see **Template saved.**, the panel closes and the list
refreshes.
## Every field on the form
The form is the same for a new template (panel title **New WhatsApp template**) and for an existing one (panel title
**Edit WhatsApp template**). Only the account owner can open it and save it.
| Field | What it does | Allowed values and limits | Default | Who can change it |
| - | - | - | - | - |
| **Purpose** | Decides when RuleInside uses the template. | **Immediate follow-up**, **Event follow-up**, **Re-engagement (campaigns)** or **Manual send**. See [The four purposes](#the-four-purposes). | **Re-engagement (campaigns)** | Owner |
| **Kind** | A normal message, or a request for permission to place a WhatsApp call. | **Message** or **Call permission**. | **Message** | Owner |
| **Template name** | The template's name exactly as it appears in Meta's WhatsApp Manager. The example in the box is `order_confirmation`. | Required. Up to 512 characters. Spaces at either end are removed when you save. | Empty | Owner |
| **Language code** | Meta's language code for the approved template. The example in the box is `en_US`. | Required. Up to 20 characters. Spaces at either end are removed when you save. Must match Meta exactly: `en` and `en_US` are two different templates at Meta. | Empty | Owner |
| **Meta category** | The category Meta gave the template. It decides whether the weekly marketing limit and the marketing quiet hours apply when the template is sent. | **Marketing**, **Utility** or **Authentication**. | **Marketing** | Owner |
| **Send from (WhatsApp number)** | Which of your WhatsApp Business numbers sends this template. Leave it on automatic unless you run more than one number. | **Automatic (account default)**, or one of your WhatsApp numbers, shown as the number followed by the business account name in brackets. | **Automatic (account default)** | Owner |
| **Template variables** | One line for each numbered blank in the template body, in order, saying what fills it. | Up to 20 lines. Every line needs a value. See [Template variables](#template-variables). | No lines | Owner |
| **Enabled** | Only an enabled template is used for sends and campaigns. | On or off. | On | Owner |
RuleInside does not check the name, language or number of variables against Meta when you save. A mistake shows up
afterwards in the **At WhatsApp** column, or as a failed send. The form's own reminder says it: the name, language
and parameter count must match Meta exactly.
The **Send from (WhatsApp number)** field appears only when your list of WhatsApp numbers can be loaded. If you do
not see it, open **Phone numbers** and check that your WhatsApp number is listed and connected.
### The four purposes
* **Immediate follow-up**: sent automatically right after a shopper who messaged your store on WhatsApp leaves their
details in that chat (for example, to send a promised discount code). It is used only when automatic WhatsApp
follow-up is switched on for your store, which it is not by default, and only if this template is enabled.
* **Event follow-up**: sent by an [automation](/automations) rule of the event follow-up type.
* **Re-engagement (campaigns)**: the template your WhatsApp [campaigns](/campaigns) send.
* **Manual send**: for sending one message yourself from [WhatsApp send](/whatsapp-send).
### Message and Call permission
Follow-ups, campaigns and manual sends all use **Message** templates. You can register a **Call permission**
template, but it cannot be chosen on **WhatsApp send**, which says call-permission templates can't be used to send a
message.
## Template variables
A template body can contain numbered blanks such as `{{1}}` and `{{2}}`. Each line under **Template variables** fills
one blank: the first line fills `{{1}}`, the second fills `{{2}}`, and so on. Each line shows the blank's number, a
source, a value box and a red **Remove** button.
| Source | What fills the blank | Value to type | Default |
| - | - | - | - |
| **Contact field** | A detail from the saved contact whose number you are messaging. | `name`, `email`, `phone`, or the exact name of a custom field on the contact. | A new line starts as **Contact field** with `name`. |
| **Send-time value** | A value typed at the moment of sending. On **WhatsApp send** you get one box for each name you choose here. | A short name for the value, for example `order_number`. | Empty |
| **Fixed text** | The same text on every send, for example a discount code. | The text itself. | Empty |
Rules for variables:
* **Order matters.** Removing a line renumbers the lines after it.
* **Only the body.** RuleInside fills the numbered blanks in the template body. It does not fill anything in the
header or the buttons.
* **No empty lines.** If any line has no value, **Save template** stops with **Fill in every variable (or remove the
empty ones).** and nothing is saved.
* **Changing a line's source clears its value.** A line switched back to **Contact field** returns to `name`.
* **The count must match Meta.** RuleInside cannot see the template body at Meta, so it cannot check that you added
one line per blank. Meta refuses a send when the count is wrong.
* **Up to 20 lines.** At 20, **Add variable** is greyed out and the form says **Maximum of 20 variables reached.**
* **No blanks at all?** Leave the list empty. The form then says the template has no numbered blanks in its body.
When a send happens, line breaks, tabs and repeated spaces in each value are turned into single spaces. If a value
comes out empty (for example the contact has no name saved), the message is not sent: it is recorded as **Failed**
with the detail `unresolved template params at index [n]`. The number counts from 0, so `index [0]` means `{{1}}`.
Fill in that detail on the contact, or change the line's source.
## Buttons on the form
| Button | What happens |
| - | - |
| **Save template** | Checks the form, then saves. It reads **Saving…** while it works. On success you see **Template saved.** and the panel closes. If saving fails, a red message explains why (or says **Could not save the template.**) and the panel stays open with your entries. |
| **Cancel** | Closes the panel without saving. Clicking outside the panel or its close button does the same. |
| **Add variable** | Adds a variable line at the end. Nothing is saved until **Save template**. |
| **Remove** | Removes that variable line. Nothing is saved until **Save template**. |
## One template per purpose and kind
Each purpose can hold one **Message** template and one **Call permission** template. A second template for the same
purpose and kind does not sit beside the first: it replaces it.
When the **Purpose** and **Kind** you pick already belong to another template, an amber warning appears at the top
of the form: "A template for this purpose and kind already exists - saving will replace it, not add a new one."
It does not stop you. If you save, the other template's registration is replaced.
The new-template form starts on **Re-engagement (campaigns)** and **Message**, so if you already have a
re-engagement message template you see this warning as soon as the panel opens. Change the purpose before saving
unless you mean to replace it.
## Editing a template
Click **Edit** on a row. The **Edit WhatsApp template** panel opens with that template's settings filled in. Change
what you need and click **Save template**.
**Changing the Purpose or Kind saves a copy and leaves the original.** The template is saved under the new purpose
and kind (replacing whatever was there), and the original row stays exactly as it was. To move a template to another
purpose, save it under the new purpose, then delete the old row.
### Enabling and disabling
Turn **Enabled** off and save to stop RuleInside using a template without deleting it. The **Status** column then
shows **Disabled**. While no enabled message template exists for a purpose:
* manual sends for that purpose are refused;
* automatic follow-ups for that purpose send nothing;
* campaigns that rely on it do not send.
Turn it back on and save to use it again.
## Deleting a template
Click **Delete** on a row. A dialog titled **Delete this template?** explains that the template will be removed and
that any follow-up or campaign relying on it will stop sending until you add a replacement.
* **Keep template** closes the dialog. Nothing changes.
* **Delete template** removes the registration. It reads **Deleting…** while it works, and you see **Template
deleted.** when it is done.
Deleting cannot be undone in the portal. To use the template again you register it again and re-enter its purpose,
kind, category, number and variables. The template itself is **not** deleted at Meta: it stays in WhatsApp Manager.
If deleting fails, a red message explains why (or says **Could not delete the template.**). If the template had
already been removed, for example by another owner, the message is **WhatsApp template not found.**
## Template statuses
Each row has two separate statuses.
### Status (your switch)
| Status | Meaning |
| - | - |
| **Enabled** (green) | RuleInside uses this template for its purpose. |
| **Disabled** (grey) | RuleInside does not use it. It cannot be picked for a manual send, and follow-ups and campaigns for that purpose do not send. |
### Status at WhatsApp
For a number connected through Meta (it shows the **Connected through Meta** badge on **Phone numbers**), each time
the page opens RuleInside asks Meta for your templates and matches each row by its **name and language together**.
| Status | Colour | Meaning |
| - | - | - |
| **Approved** | Green | Meta has approved the template. |
| **Pending review** | Amber | Meta is still reviewing it. |
| **Rejected** | Red | Meta rejected it. |
| **Paused** | Red | Meta paused it. |
| **Disabled** | Red | Meta disabled it. This is Meta's status, not the same as **Disabled** in the **Status** column. |
| **Not on your WhatsApp account** | Red | Meta's full list has no template with this exact name and language. Usually a typo, the wrong language code, or a template approved on a different WhatsApp account from the one now connected. |
| **Could not check** | Grey | Meta could not be read just then (it did not answer in time, refused, or the connection needs attention), or your template list is very long and this template was not in the part that was read. Refresh later. |
| **Not checked** | Grey | Your WhatsApp number is not connected through Meta, so template statuses cannot be read. It does not mean anything is wrong. |
| Any other word | Red | A status Meta reported that RuleInside has no label for (for example ARCHIVED or DELETED), shown as Meta sent it. Only an approved template can be sent, so it is shown as a problem and never treated as approved. |
When Meta's category for a template differs from the **Meta category** you chose, a small line under the status
says so, for example **WhatsApp lists it as UTILITY**. Edit the template and pick the category Meta shows, so the
marketing limit and quiet hours apply correctly.
**At WhatsApp** is there to warn you; it does not stop a send. A template shown as **Rejected** or **Not on your
WhatsApp account** but still **Enabled** is still tried, Meta refuses it, and the send shows **Failed** in the send
log. Fix or disable the template when you see a red status.
## How the category affects sending
The **Meta category** you choose here decides two rules for every send of the template:
* **Weekly marketing limit**: a contact receives at most 4 Marketing WhatsApp messages in any rolling 7 days. Utility
and Authentication templates do not count and are not limited.
* **Marketing quiet hours**: a Marketing template is sent only during the calling hours where the person is, and
waits until then otherwise (an immediate follow-up sent within 24 hours of the shopper's chat is the exception).
Utility and Authentication templates do not wait.
Choose the category Meta gave the template. Every send also passes the consent, opt-out and do-not-contact checks
described in [Why a send is suppressed](/whatsapp-send#why-a-send-is-suppressed).
## STOP and START
A customer who replies **STOP** is opted out of WhatsApp messages and calls from your store, and every later send to them is
suppressed, whatever template you use. Only the customer can opt back in, by replying **START**. See
[STOP and START](/whatsapp#stop-and-start).
## Who can do what
* **Owner**: register, edit, enable, disable and delete templates.
* **Member**: sees the list and the statuses. In place of **New template** they see a note that only account owners
can manage templates, and the **Edit** and **Delete** buttons are greyed out.
## Common questions
No. Write it, add any header, buttons and footer, and get it approved in Meta's WhatsApp Manager. Then register
it here with its exact name and language. To change the wording, edit it at Meta.
In Meta's WhatsApp Manager. RuleInside fills only the numbered blanks in the template body.
Meta has no template with that exact name and language on the connected account. Check the spelling and the
language code (`en` is not `en_US`), and check that the template was approved on the WhatsApp account that is
connected now, not an older one. Edit the row to fix it.
Your WhatsApp number is not connected through Meta, so RuleInside cannot read template statuses. Nothing is
wrong; just make sure the name and language match WhatsApp Manager exactly.
Meta classifies the template differently from the category you chose. Edit the template and choose the category
Meta shows, so the weekly limit and quiet hours apply correctly.
There is one template per purpose and kind, so saving a second one for the same pair replaces the first (the
form warns you before you save). To keep both, register the old one again under another purpose.
That is expected: changing **Purpose** or **Kind** saves a copy under the new pair and leaves the original.
Delete the original if you no longer want it.
You can register a call permission template, but follow-ups, campaigns and manual sends all use **Message**
templates, and **WhatsApp send** cannot use a call permission template.
It appears only when your list of WhatsApp numbers loads. Open **Phone numbers** and check your WhatsApp number
is there and connected.
No. Only the registration in RuleInside is removed. The template stays in WhatsApp Manager and you can register
it again. Follow-ups and campaigns for that purpose stop until you register a replacement.
No. WhatsApp requires a pre-approved template to start a conversation, so RuleInside cannot send free-form text.
No. Members can see the list and statuses, but only the account owner can change them.
# Writing your agent's greeting and instructions
Source: https://docs.ruleinside.com/writing-instructions
How to write a first message and instructions that work on a call: what to include, what to leave to RuleInside, examples you can copy, and how to test each change.
This page helps you write the words your agent works from. When you finish, you will be able to:
* write a greeting that tells callers who they are talking to and invites them to speak,
* write **Additional instructions** that give your agent your store's tone and priorities without repeating or
contradicting what RuleInside already handles,
* test a change and check it against real conversations.
Plan about half an hour for a first version, then a few short rounds of testing.
## Where you write them
Everything on this page is written in the agent editor: click **Agents** in the sidebar. The greeting is the
[First message](/agent-settings#first-message) card and the instructions are the
[System prompt](/agent-settings#system-prompt) card. What you can write depends on the agent's
[edit level](/agent-settings#edit-levels), shown as a badge in the editor header.
| Edit level | What you can write | Limit |
| - | - | - |
| **View only** | Nothing. You can read the current instructions. | |
| **Name, voice & greeting** | The **First message** and **Personalization variables** | 1,000 characters for the first message |
| **Instructions** | Everything above, plus **Additional instructions**, **Who speaks first**, the **Recording notice** sentence and, when transfers are on for your account, **When to transfer** for each number | 8,000 characters for Additional instructions; 500 for the recording sentence and for each **When to transfer** |
| **Full control** | Everything above, plus the whole **System prompt**, which takes the place of Additional instructions | 20,000 characters |
Only account **owners** can change an agent. Members can read every setting but cannot edit. Agents you create
yourself always start at **Full control**; your main agent keeps the level RuleInside set for it.
## What is already handled for you
Your agent arrives with instructions and store tools set up by RuleInside. At the **Instructions** level the
System prompt card says so: "Platform behaviour (memory, store lookups, call handling) stays managed." Your
Additional instructions are added in their own section, and the rest of the prompt is never touched.
Before you write anything, open the **System prompt** card and click **Show** under **Current agent instructions
(managed)**. Read what is already there. Your job is to add what only you know: your store, your products, your
tone. You do not need to describe the things below, and you should not contradict them.
| Already handled | How it works | What not to write |
| - | - | - |
| **Order questions** | For a customer logged in to your store: the status of their own orders, and the tracking number when your store records one. Someone not logged in gets an order's details only by giving its order number together with the billing email and postcode on that order. | Instructions that promise order details "for any order", or on less than those checks. |
| **Product search** | Searches your catalogue by meaning, not only by exact words. | A product list pasted into the instructions to "search". |
| **Coupon codes** | Tells the shopper whether a coupon code is valid. | Instructions that promise a discount the store has not set up. |
| **Refund requests** | The agent records the request and tells the shopper: "Your refund request has been recorded and sent to the store owner to review. They will get back to you about it." A person approves every refund; nothing is refunded automatically. Only shoppers your store has recognised can make a request. | "Approve refunds under 20 euros", or any promise of a refund, an amount or a date. See [Refund requests](/refunds). |
| **Recording objections** | If a caller says they do not want to be recorded, that call's recording and transcript are deleted. With the recording notice said after the first exchanges, the agent says someone from the team will get back to them and ends the call. | Your own script for what to do when someone objects to recording. |
| **The store tools** | Order look-up, product search and coupon check are marked **Managed** on the [Tools](/tools) page and always stay on the agent. | Instructions to "turn off" or replace them. |
Some behaviour is set by a setting, not by words. Call length, silence, interruptions, the language and the time
zone all have their own fields in [Agent settings](/agent-settings). We recommend using the setting rather than
writing an instruction for it: for example, set **Maximum call length (minutes)** under **Call limits** instead of
writing "keep calls under ten minutes".
## The greeting
The greeting is the **First message**: "The first thing customers hear." It is up to 1,000 characters, and it is
required when **Agent speaks first** is chosen.
### What a good greeting contains
1. **That the caller is talking to an AI assistant.** Always include this. The portal does not stop you saving a
greeting without it, but always include it.
2. **That the call may be recorded**, when **Recording notice** is set to **In the opening line**. If you choose
**After the first exchanges** instead, take the recording sentence out of the greeting, or it will be said twice.
The greeting must still say the caller is talking to an AI assistant.
3. **Your store's name**, so the caller knows they reached the right place.
4. **One open question** that hands the turn to the caller, such as "What can I help you with today?"
### Keep it short for the ear
A greeting is heard, not read. Some advice:
* Aim for two or three short sentences, about ten seconds when you read it aloud. Read it aloud before you save.
* End on the question, so it is clear that it is the caller's turn to speak.
* Leave out web addresses, symbols, abbreviations and opening hours. They are hard to follow by ear, and are better
given later, if the caller asks.
* If you want the recording notice always heard in full, switch on **Callers cannot interrupt the greeting** under
**Conversation behaviour**.
* Write the greeting in the language your agent speaks.
### Example greetings
Replace the parts in angle brackets with your own details.
A fashion store:
```text theme={null}
Hi, thanks for calling . I'm the store's AI assistant, and this call may be recorded. Are you looking for something new, or can I help with an order?
```
An electronics store:
```text theme={null}
Hello, you've reached . I'm an AI assistant and this call may be recorded. Tell me what you're looking for, and I'll help you find the right model.
```
A food store:
```text theme={null}
Hi, welcome to . You're speaking with our AI assistant, and this call may be recorded. What can I help you with today?
```
With **Recording notice** set to **After the first exchanges**, the same greeting keeps the AI part and drops the
recording part:
```text theme={null}
Hi, welcome to . You're speaking with our AI assistant. What can I help you with today?
```
The recording sentence then goes in **What the agent says** (up to 500 characters, said word for word), for
example "Just so you know, this call may be recorded for quality purposes."
### Greeting a customer by name
[Personalization variables](/agent-settings#personalization-variables) are values your storefront passes in when a
call starts, such as the customer's first name. Write a variable's name in double curly braces, for example
`{{first_name}}`. When no value is passed in, the default you set is used instead.
Choose a default that still reads naturally. With the default `there`:
```text theme={null}
Hi {{first_name}}, welcome back to . I'm the store's AI assistant, and this call may be recorded. How can I help today?
```
A known customer hears "Hi Laura, welcome back…", and everyone else hears "Hi there, welcome back…". If "welcome
back" would sound odd to a first-time visitor, drop it:
```text theme={null}
Hi {{first_name}}, thanks for visiting . I'm an AI assistant and this call may be recorded. What are you shopping for today?
```
Write the variable's name exactly as it appears in your list of variables. **Save variables** saves that list
straight away; the first message saves with **Save changes**.
The website assistant has its own **Returning-shopper greeting**, which uses `{name}` with single braces and has its
own rules. See [The returning-shopper greeting](/website-assistant#the-returning-shopper-greeting).
### Who speaks first
With **Agent speaks first**, the greeting is said as soon as the call connects. With **Caller speaks first**, the
agent waits for the person to answer ("Hello?"), then says the greeting. Either way, the greeting is the same text
and needs the same AI disclosure. **Who speaks first** needs the **Instructions** level.
## Writing Additional instructions
**Additional instructions** take up to 8,000 characters: "Added to your agent's instructions. Leave blank to
remove." You do not need to use all of them. A clear page of instructions usually does more than a long one.
### A simple structure
We recommend five short parts, in this order. Plain labels in capitals, as in the example further down, keep it
easy for you to find things later.
1. **Who you are.** The store's name, what it sells and who buys from it.
2. **Tone.** How the agent should sound, in a sentence or two.
3. **What to help with.** The most common reasons people contact you, in order of importance, with the facts the
agent needs for each.
4. **What to avoid.** The topics that are out of scope, said once, in one place.
5. **When to hand over.** What to do when the agent cannot help: your support email, or a transfer when transfers
are on.
### Write for the ear
Everything the agent says is spoken. Some advice:
* **Short sentences.** Ask for replies of one or two sentences, then let the customer talk.
* **One question at a time.** "What size do you usually wear?" works better on a call than three questions in one
breath.
* **Two options, not ten.** Ask the agent to offer two choices and then ask if the customer wants to hear more.
* **No lists or formatting in what it says.** Tell the agent to speak in sentences and never to read out bullet
points, symbols or web addresses.
* **Show how to say numbers.** Give one example of how a price, a weight or a phone number should sound, such as
"say 24.99 euros as twenty-four ninety-nine". If you give your phone number or email address, write them the way
they should be spoken.
* **Help with names.** If a brand or product name is easy to mishear, add it to **Words to recognise** under
**Conversation behaviour**, so the agent catches it when a customer says it.
### Be specific
"Be helpful" and "be professional" give the agent nothing new to do. Say what helpful looks like in your store.
| Instead of | Try |
| - | - |
| Be helpful and friendly. | Sound like a relaxed shop assistant: warm, brief, everyday words. Answer the question first, then offer one useful next step. |
| Know our shipping policy. | For shipping costs and delivery times, answer from the Shipping and returns document. If it does not cover the question, say so and give our support email. |
| Recommend products. | When someone describes what they need, suggest at most two products that fit, and say in one sentence why each fits. Then ask which sounds better. |
| List the sizes: S, M, L, XL. | When asked about sizes, say the range in a sentence, such as "we have it from small to extra large", then ask which size they usually wear. |
| Prices: €24.99, €39.00. | Say prices the way a person would: "twenty-four ninety-nine", "thirty-nine euros". |
| Handle angry customers well. | If someone is upset, say you understand in one short sentence, then deal with the problem. Do not keep apologising. |
| Upsell whenever possible. | When someone has chosen a camera, offer a matching memory card once. If they say no, do not offer it again. |
| Never discuss politics. Never talk about other shops. Never give opinions on the news. | Keep the conversation on our products, orders and store policies. If asked about anything else, say politely that you can only help with the store, and ask what they need. |
### Do not add a "never" rule after every bad call
When a conversation goes wrong, it is tempting to add a line such as "never do that again". After a few weeks the
instructions become a long list of exceptions that is hard to keep consistent: rules start to overlap, and some
contradict each other or the managed instructions.
Instead, find the instruction that left the gap and rewrite it. For example, suppose the agent recommended a
grinder to someone who had just said they buy ground coffee. Rather than adding "never suggest a grinder to people
who buy ground coffee", rewrite the original line:
```text theme={null}
Before: Recommend a grinder with every coffee order.
After: Suggest a grinder only to customers who buy whole beans and do not have one yet.
```
One clear boundary is fine, such as "Stay on coffee, brewing and the shop." It is a long list of narrow
prohibitions that we recommend avoiding.
### Referring to your documents
Facts that change, such as shipping rates, return windows, opening hours or a size guide, are best kept in
[Knowledge](/knowledge) documents rather than typed into the instructions. When they change, you update one
document instead of searching your instructions.
In the instructions, say **when** to use a document, not what it says:
```text theme={null}
For questions about delivery costs, delivery times and returns, answer from the Shipping and returns document.
```
Adding or changing documents needs **Full control** on the agent. On your main agent below Full control, contact
RuleInside to add a document for you. A document in **Prompt** mode is always given to the agent; one in **Auto**
mode is looked up when it is relevant.
### Describing when to transfer
If RuleInside has switched transfers on for your account, each number under
[Transfer to a person](/agent-settings#transfer-to-a-person) has its own **When to transfer** box (up to 500
characters, required). Write the transfer conditions there, and be specific:
```text theme={null}
The caller asks to speak with a person, wants to change an order that has already shipped, or is a business asking about wholesale prices.
```
Some advice:
* Keep the conditions in **When to transfer**, and keep your Additional instructions consistent with them. Two
different versions of the same rule are hard to maintain.
* Transfers work on phone calls only. For conversations on your website, tell the agent what to offer instead, such
as your support email.
* If the card says "No transfer numbers. The agent never hands a call over.", do not write instructions that promise
a transfer.
### At Full control: editing the whole prompt
At **Full control** the System prompt card shows the whole prompt in one box, up to 20,000 characters, instead of
Additional instructions. All the advice above still applies. In addition:
* **Keep RuleInside's sections.** If your agent's prompt contains RuleInside's platform sections, they must still be
there when you save, or the save is refused with "Your system prompt must keep its platform sections (…).
Restore them and save again." A blank agent you created has none.
* **Copy the prompt before a big edit.** Paste it into a document on your computer first, so you can put it back.
* **Leave alone what you did not write.** If you are not sure what a line does, keep it and contact RuleInside.
* The greeting (when the caller speaks first) and the recording sentence (when said after the first exchanges) are
kept in their own sections. They are not shown in this box and are kept automatically when you save.
## A complete example
Here is a full set of Additional instructions for a fictional WooCommerce coffee shop. It is about 2,900
characters, well under the 8,000 limit. Copy it, then replace every detail with your own. It assumes a Knowledge
document called "Shipping and returns" is attached to the agent.
```text theme={null}
WHO YOU ARE
You are the voice assistant for Fernhill Coffee Roasters, an online shop that roasts and sells coffee beans and brewing equipment. We roast in small batches every Monday and Thursday and ship from one warehouse. Our customers range from beginners with their first French press to people who weigh every dose.
TONE
Warm, relaxed and knowledgeable, like a good barista on a quiet morning. Use everyday words and avoid jargon unless the customer uses it first. Keep most replies to one or two short sentences, then let the customer talk. Ask one question at a time. Speak in full sentences: never read out lists, bullet points, symbols or web addresses. When there are several options, give two and ask if they would like to hear more.
WHAT TO HELP WITH, IN THIS ORDER
1. Choosing coffee. Ask how they brew: espresso, filter, French press, moka pot or cold brew. Then ask whether they like chocolatey and smooth, or bright and fruity. Suggest one or two coffees that fit, and say in one sentence why.
2. Choosing equipment. For beginners, start with the simplest option that fits their budget. Suggest a grinder only to customers who buy whole beans and do not have one yet.
3. Brewing help. Give one clear tip at a time and check it helped before giving another. For filter coffee, our starting recipe is fifteen grams of coffee to two hundred and fifty grams of water.
4. Shipping and returns. For delivery costs, delivery times and returns, answer from the Shipping and returns document. If it does not cover the question, say so and give our support email.
5. Orders and refunds. Use the store's built-in order and refund handling as usual. Never promise a refund, a refund amount or a date.
HOW TO SAY THINGS
Say prices the way a person would, for example "fourteen ninety-five" for 14.95 euros. Say weights in words, for example "two hundred and fifty grams". Our three house coffees are Alba, Campo and Notturno; say them as written. Our decaf is called Notturno Decaf.
WHAT TO AVOID
Stay on coffee, brewing and the shop. If someone asks about something else, say politely that you can only help with Fernhill Coffee, and ask what they would like to know. Do not give medical advice about caffeine: mention Notturno Decaf, and suggest they ask their doctor about anything medical. Do not guess stock levels, roast dates or delivery dates. If you cannot find the answer, say you are not sure.
WHEN YOU CANNOT HELP
Our team answers email Monday to Friday, nine in the morning to five in the afternoon. Give the support address, spoken as "hello at example dot com", and say someone will reply within one working day. If the customer is upset, say you understand in one short sentence before giving the address.
ENDING THE CONVERSATION
When the customer has what they need, ask once if there is anything else. If not, thank them for choosing Fernhill Coffee and say goodbye.
```
## Testing your changes
A change is only finished when you have heard it work. We recommend this routine:
Click **Save changes** in the bar at the bottom of the editor and wait for "Agent updated." Closing a panel does
not save it.
Click **Preview** in the editor header. Preview talks to the real agent with its **saved** settings. Start with
**Text** to check the words, then try **Voice** to hear the greeting and the pace. See [Testing](/testing).
Keep a short list of five or six real questions, such as "what's your cheapest grinder?" or "how long does
delivery take?", and ask the same ones after every change, so you can compare.
After real customers have talked to the agent, open [Calls](/conversations) and read a few transcripts. The
transcript also shows the actions the agent took and when it checked your knowledge base, which tells you
whether it used the document you expected.
If you change the tone, the greeting and three rules at once, you cannot tell which change made the difference.
Make one change, test it, then make the next. Keep a note of what you changed and when.
Preview does not identify you as a customer, so store tools answer as they would for a guest. To check order
answers for logged-in customers, read real conversations in **Calls**.
## Common questions
First check that the change was saved: you should have seen "Agent updated." Preview uses the saved settings
only. Then click **Show** under **Current agent instructions (managed)** and look for anything that says the
opposite of your instruction. If your instruction is about orders, refunds or recording, it is already handled
for you (see [What is already handled for you](#what-is-already-handled-for-you)). Otherwise, rewrite the
instruction to be more specific, with an example of what the agent should say, and test again.
Check your instructions and documents for bullet points, symbols and web addresses the agent may repeat. Add a
line such as "Speak in full sentences and never read out symbols or web addresses", and give one spoken example
for prices or phone numbers. At **Full control**, the **Text normalisation** setting under **Voice** decides how
numbers and symbols are turned into speech; contact RuleInside before changing it if you are not sure.
Ask for replies of one or two sentences, and for two options at a time rather than full lists. Make sure the
instructions do not ask for long explanations elsewhere. Callers can also cut in: with **Callers can interrupt
the agent** on (under **Conversation behaviour**), the agent stops talking as soon as the caller speaks.
Order look-up is one of RuleInside's store tools, marked **Managed**, and you cannot change or remove it. It
answers a customer logged in to your store about their own orders, and anyone else only when they give the
order number together with the billing email and postcode on that order.
Your instructions can set the tone around it, but they cannot widen who gets order details. If you need
something changed, contact RuleInside.
Either you are a member rather than an owner, or the agent's edit level does not include that field.
**Additional instructions** need the **Instructions** level; the whole **System prompt** needs **Full control**;
a **View only** agent cannot be changed at all. See [Edit levels](/agent-settings#edit-levels).
You chose **After the first exchanges** but the greeting still contains the recording sentence. Take it out of
the greeting and keep the AI disclosure.
No value was passed in for that variable, so its default was used. Check the default under **Personalization
variables**, and check that the name in the greeting matches the variable's name exactly.
Not refunds: a person approves every refund, and the agent only records the request. For discounts, the agent
can tell a shopper whether a coupon code is valid. We recommend not writing instructions that promise discounts
your store has not set up.
Empty the box and click **Save changes**. Saving it empty removes your section; the managed instructions stay
as they are.
## Next steps
Every field in the agent editor, including the first message and the system prompt.
Try your changes with Preview before customers hear them.
Add the documents your agent answers from.
Read real conversations and see where the agent can improve.
# Your account
Source: https://docs.ruleinside.com/your-account
Your personal profile, signing in safely, and how to sign up, reset a password, accept an invitation and create an organization.
Your account is **you**: your name, picture, email, password and sign-in security. It is separate from your
organization, and it applies in every organization you belong to. Only you can change it, and nobody can change it
for you from their own portal.
## Your Profile
Open **Settings** from the left sidebar, then the **Profile** tab. The cards appear in the order below.
### Your Profile Picture
Click **Upload a Profile Picture** and choose an image file. A new image replaces the old one; clearing it removes
your picture. You see "Updating profile..." and then "Profile successfully updated".
### Your Name
Type your name in the field and click **Update Profile**. It must be 2 to 100 characters. This is the name your
colleagues see in the **Name** column on the [Members](/members) page. Without a name, they see the part of your
email before the `@`.
The portal is in English only, so there is no language setting.
### Update your Email
Fill in **Your New Email** and **Repeat Email**. Both must be valid, must match, and must differ from your
current email.
A green box confirms: "We sent you an email to confirm your new email address."
Open your inbox at the new address and click the link.
Your email does not change until you click the confirmation link. Until then, keep signing in with your current
address.
If you use two-factor sign-in but did not use it when you signed in this time, you may see "Please sign-in with
your current multi-factor authentication to perform this action". Sign out, sign back in with your code, and try
again.
### Update your Password
Fill in **New Password** and **Repeat New Password**, then click **Update Password**. A green box says "Your
password has been successfully updated!"
* The password must be 8 to 99 characters, and the two fields must match.
* There is no "current password" field: being signed in is the proof that it is you.
* It cannot be the same as your current password.
* "This password is too weak. Use at least 8 characters and mix letters, numbers and symbols." means you should
choose a stronger one.
* If a yellow **Reauthentication Required** box appears, sign out, sign in again, and then change the password.
### Multi-Factor Authentication (two-factor sign-in)
Two-factor sign-in adds a second check after your password: a 6-digit code from an authenticator app on your phone
or from a password manager. The code changes every few seconds, so someone who steals your password still cannot
get in.
In the **Multi-Factor Authentication** card, click **Setup a new Factor**.
In **A memorable name to identify this factor**, type a name you will recognise later, for example "iPhone".
Click **Set factor name**.
Scan the QR code with your authenticator app or password manager.
Type the 6-digit code the app shows and click **Enable Factor**. You see "Factor successfully enrolled".
From then on, every sign-in asks for a code. Only authenticator-app codes are offered; there is no text-message
option.
Once you have a factor, the card lists it with its **Factor Name**, **Type** and **Status**. To remove one, click
its **X** button (**Unenroll this factor**) and confirm with **Yes, unenroll factor**. You can no longer use that
factor to sign in.
If you have no factor, your dashboard shows a reminder, "Add a second step to your sign-in", with a link **Set up
two-step sign-in**. It is a reminder, not a requirement. You can dismiss it; it may show again on another device.
Use a password manager that keeps your codes in sync across devices. If you lose your phone, you still have the
codes elsewhere.
### Passkeys
If your Profile shows a **Passkeys** card, you can sign in without a password using your device's fingerprint,
face unlock or a security key.
* Click **Add a passkey** and follow your browser's prompt. You see "Passkey successfully registered".
* Each passkey shows its name and "Added on" with the date.
* To delete one, click its **X** button (**Delete this passkey**) and confirm with **Yes, delete passkey**. You can
no longer sign in with it.
* Once you have a passkey, the sign-in page offers **Sign in with Passkey**. A passkey signs you in to an account
that already exists; it cannot create a new account.
### Linked Accounts
**Linked Accounts** shows the ways you can sign in.
* **Sign-in methods linked to your account** lists each method, for example "Email", with its address. When you
have more than one, **Unlink Account** removes one. Unlinking cannot be undone.
* **Available Methods** lists what you can add:
* **Email & Password** adds a password to your account. It is how someone who joined through an invitation
link sets a password. It is shown when you have an email login but did not sign in with a password this time.
* **Set Email Address** adds an email, and appears only when your account has none.
* Sign-in with Google or other social accounts is not available.
### Danger Zone: Delete your Account
**Delete your Account** permanently deletes your login and removes you from every organization you belong to.
**Deleting your account is refused while you are the primary owner of an organization.** Transfer ownership
first, on the [Members](/members#transferring-ownership) page.
**Deleting your account does not cancel a subscription.** If you want to
stop paying, cancel your plan on [Billing](/billing#cancelling-your-plan) first.
A dialog with the same title opens.
We email you a 6-digit code. Type it in to continue.
Click **Delete your Account**. You see "Deleting account. Please wait...".
You are signed out and receive a goodbye email. **This cannot be undone.**
## Signing up
You sign up with an email and a password. Sign-in with Google or other social accounts is not available.
1. Open the sign-up page, **Create an account**.
2. Fill in **Email**, **Password** (8 to 99 characters; the eye icon shows what you typed) and **Repeat password**.
3. Tick "I accept the Terms of Service and Privacy Policy". This is required.
4. Click **Sign up with Email**.
5. You see **We sent you a confirmation email.** Open it and click the link to verify your address.
If you were invited, use the link in your invitation email instead: see
[Accepting an invitation](#accepting-an-invitation).
## Verifying your email
Clicking the link in the confirmation email signs you in. You land in your organization or, if you do not have one
yet, on the page to [create an organization](#creating-an-organization).
If the link does not work, you see **Authentication Error** with one of these messages:
| Message | What to do |
| - | - |
| "The email link has expired. Please try again." | Use the **Resend link** form on the same page. You see **Check your email!** when a new link is on its way. |
| "It looks like you're trying to sign in using a different browser than the one you used to request the sign in link. Please try again using the same browser." | Open the link in the browser you used to sign up. |
| "Sorry, we could not authenticate you. Please try again." | Click **Sign In** and try again. |
## Signing in
1. On **Sign in to your account**, enter your email and password.
2. Click **Sign in with Email**.
3. If you use two-factor sign-in, the page **Verify your account** asks for the 6-digit code from your app. With
more than one factor, first choose which one to use. Click **Submit Verification Code**.
After signing in you go to the page you were trying to open, or to your dashboard.
Messages you may see:
* "The credentials entered are invalid": the email or password is wrong.
* "Please confirm your email address before signing in": click the link in your confirmation email first.
* **Invalid Verification Code**: the two-factor code was wrong or expired. Enter the current code.
## Resetting a forgotten password
On the sign-in page, click **Forgot Password?**. On **Reset Password**, type your **Email Address** and click
**Reset Password**.
You see "Check your Inbox! We emailed you a link for resetting your Password." Click the link in that email.
Type the new password twice (8 to 99 characters) and click **Reset Password**. You see "Your password has been
successfully updated!" and go to your dashboard.
If it fails you see **Something went wrong**, often with the reason, for example that the password is too weak or
the same as the current one. Click **Retry**.
## Accepting an invitation
When an owner or member invites you, you receive an email with a link.
1. Click the link. You are signed in automatically. If you do not have an account yet, one is created for your
invited address.
2. The page **Join** followed by the organization's name appears, with its logo if it has one.
3. Click **Continue as** followed by your email. You see "Joining organization...".
4. If your account is new, the page **Link account to sign in** asks you to add a sign-in method. Choose **Email &
Password** to set a password, then click **Continue**.
If you skip step 4, the portal asks **No authentication method added**. You can continue, but your account then
has no password. Next time, use **Forgot Password?** on the sign-in page, or add one later under **Profile**,
**Linked Accounts**.
Things that can go wrong:
* **Invite not found or expired**: the invitation is more than 7 days old, was removed or was already used, or you
are signed in with a different email from the one that was invited. Ask the person who invited you to renew it,
or to invite the email you sign in with.
* **A "page not found" message straight after clicking the link**: usually the invitation has expired or been
removed. Ask for it to be renewed or sent again.
* **You are signed in as someone else**: under the button, use **Sign in with a different account**, then sign in
with the invited address.
* **You are already a member**: you go straight to your dashboard.
## Creating an organization
If you belong to no organization, the portal takes you to **Create your organization** after you sign in.
1. Type an **Organization Name**. The rules are the same as for renaming: 2 to 50 characters, no special
characters, not `settings` or `billing`. For names in other scripts, an **Organization URL** field also
appears. See [Organization settings](/organization-settings#organization-name).
2. Click **Get Started**.
You become the organization's **primary owner**, and the portal opens its dashboard. Your assistant is set up by
our team after you choose a plan on [Billing](/billing).
If you already belong to an organization, you see **You already belong to an organization** with a button to go to
it. To create another one, open the workspace switcher, choose **Create an Organization**, type the name and click
**Create Organization**.
## Common questions
**Settings**, **Profile**, **Multi-Factor Authentication**, **Setup a new Factor**. Scan the QR code and enter
the code from your app.
The portal has no self-service way to remove a factor you cannot reach, and sign-in will keep asking for the
code. If your codes are in a password manager that syncs, use it on another device. Otherwise, contact us.
Click the confirmation link we sent to the new address. Your email changes only after that.
You are the primary owner of an organization. Transfer ownership on the [Members](/members) page first.
No. Cancel the plan on [Billing](/billing#cancelling-your-plan). Deleting your account never cancels billing.
You are already signed in, which proves it is you. If the portal needs a fresh sign-in, it tells you with
**Reauthentication Required**.
Add one under **Profile**, **Linked Accounts**, **Email & Password**, or use **Forgot Password?** on the sign-in
page.
# Your first agent
Source: https://docs.ruleinside.com/your-first-agent
A first-day checklist: look over the agent our team set up for you, try it, adjust its greeting, and get ready to go live.
By the end of this page you will have looked over the agent our team built for you, talked to it yourself, adjusted
its greeting if you want to, checked that it is on your website and your phone line, cleared your account for
calling, and invited your team. Allow **about 15 to 30 minutes**, once our team has set your agent up: most steps are
a look and a click.
**Getting started is done for you.** After you choose a plan, our team builds your first agent (we call each
assistant an *agent*), gives it a voice and sets up what it needs to search your products and help your customers.
Until it is ready, your Dashboard shows **"Setting up your AI agent"**. You can start this checklist once the
Dashboard shows your figures.
Several steps change things, and only an **owner** can make changes. If you are a **member**, you can follow every
step to look, and ask an owner for the changes. See [Who can do what](/roles).
## The checklist
Sign in on **Sign in to your account** with your email and password. The **Dashboard** opens first.
You should see four figures (**Calls**, **Talk time**, **Success rate**, **Leads captured**), a 14-day chart and
your five most recent calls. On a new account the figures may still show zero or a dash.
Read any notice at the top. A red **"Outbound calling is paused"** is expected on day one: you clear it in step
8\. See [Dashboard](/dashboard).
Click **Agents** in the sidebar. The editor of your main agent normally opens, marked with a **Managed** badge. In the
header, note the **edit level** badge: it tells you what you can change on this agent (**View only**, **Name,
voice & greeting**, **Instructions** or **Full control**).
Read the **First message** card: it is what callers hear first. Then read the **System prompt** card, which
holds how the agent is told to behave. Depending on the edit level, the instructions are shown under **Current
agent instructions (managed)** (click **Show** if they are folded away), or in one box you can edit.
See [Agent settings](/agent-settings) and [Edit levels](/agent-settings#edit-levels).
Owners only. In the editor header, click **Preview**. Choose **Text** or **Voice**, then **Start text chat** or
**Start voice call**. For voice, your browser asks for permission to use the microphone. When the status reads
**Connected**, talk or type as a shopper would. Click **End** when you are done.
Preview talks to the real agent with its saved settings. It treats you as a guest, not as a customer logged in
to your store. We suggest you ask it about a product you sell and about one of your policies.
See [Testing an agent with Preview](/managing-agents#testing-an-agent-with-preview) and [Testing](/testing).
You need the **Name, voice & greeting** level or above. In the **First message** card, edit the text (up to
1,000 characters). Keep a mention that the caller is talking to an AI assistant and that the call may be
recorded.
A bar appears at the bottom of the page. Click **Save changes**. You should see **"Agent updated."**, and the
change applies to the live agent. Run **Preview** again to hear it.
If the box is greyed out, your agent is at **View only** or you are a member. Email us the wording you want.
See [First message](/agent-settings#first-message) and [Writing instructions](/writing-instructions).
Click **Knowledge** in the sidebar. The list shows the documents your agent answers from: ours are marked
**Managed**, and yours appear below them.
Adding your own documents needs an owner and an agent at **Full control**. If you see "This agent is managed,
its knowledge is configured by RuleInside.", send us the policies or FAQs you want added. See
[Knowledge](/knowledge).
Click **Website widgets** in the sidebar. Your widget's row should show your agent's name and an **Enabled**
badge. Click the pencil to open it, and check that your store's address, starting with `https://`,
is in **Allowed domains**: the widget appears only on the websites listed there.
With the RuleInside plugin on WordPress: get it from **Settings**, **Plugin** tab, **Download plugin**. In your
WordPress admin, open the **RuleInside** menu and paste the widget's **embed id** into **Widget ID**. For
automatic plugin updates, an owner clicks **Create licence key** and pastes the key into **License Key**.
Without the plugin, copy the **Embed snippet** from the widget editor into your site's HTML.
If no widget is listed yet, an owner can create one with **New widget**. See
[Website assistant](/website-assistant) and [WordPress plugin](/plugin).
Click **Phone numbers** in the sidebar. Look at the **Inbound agent** column for your number: it should show
your agent's name. A dash means **no agent answers** calls to that number; an owner fixes it with **Assign**.
A phone campaign also needs a number that can place outbound calls. We suggest you call your number from your own phone
once, then find the call under [Calls](/conversations). See [Phone numbers](/phone-numbers).
Owners only. Open **Settings**, then the **Calling consent** tab. Choose **Country your account trades in**,
read the statement in full, tick **I have read the above and confirm it is true.** and click **Confirm**. You
should see **"Confirmed. Outbound calling is cleared for this account."**
Until an owner confirms, no calls and no WhatsApp messages are sent, even to customers who agreed to be
contacted. On the same tab, check **Calling hours and days**: by default your agent calls between 08:00 and
21:00, every day, in each customer's own time.
Each customer also needs their own consent recorded before they can be contacted. See
[Calling consent and calling hours](/calling-hours) and [Contacts](/contacts).
Click **Members** in the sidebar, then **Invite Members**. Type each person's email, choose **Member** or
**Owner**, and click **Send Invites**. You should see **"Members invited successfully"**. Each person gets an
email with a link that works for 7 days.
We suggest **Member** for people who follow what the assistant does, and **Owner** only for people who decide
how it contacts customers and who manage billing. Adding people never changes your bill. See
[Members](/members).
## Common problems on day one
Your subscription is active and our team is still building your agent. There is nothing to do: the figures
appear as soon as the agent is live. If you see **Subscribe to activate your AI agent** or **Complete payment to
activate your AI agent** instead, follow its button to [Billing](/billing).
Either you are a member, not an owner (you then see "Only account owners can change agents, ask an owner of your
team to make edits."), or the agent's edit level does not include that setting. Your main agent is often below
**Full control**. See [Edit levels](/agent-settings#edit-levels), or email us the change you want.
Only owners see **Preview** (and **New agent**). Ask an owner of your team, or ask to be made an owner on
[Members](/members).
Preview uses the agent's **saved** settings. Closing a side panel does not save it: click **Save changes** in
the bar at the bottom of the editor, wait for **"Agent updated."**, then start a new Preview.
On your website, a signed-in shopper your store recognises can hear the widget's **Returning-shopper
greeting** instead, on voice conversations. Change it in the widget editor. See
[The returning-shopper greeting](/website-assistant#the-returning-shopper-greeting).
Check these in order:
1. Your site's exact address, starting with `https://`, is in **Allowed domains**.
2. **Widget enabled** is switched on, and you clicked **Save changes**.
3. The snippet on your site, or the plugin's **Widget ID**, holds the current embed id.
4. Your account is active and on a paid plan.
5. You have not reached today's conversation limit: look for a notification in your portal.
See [Website assistant](/website-assistant).
On **Phone numbers**, the number's **Inbound agent** is a dash, so no agent answers it. An owner clicks
**Assign** and chooses an agent. See [Phone numbers](/phone-numbers).
The **New campaign** button is greyed out while your workspace is not linked to a store yet: contact us. Members
see "Only account owners can manage this: ask an owner of your team." instead of the buttons. In the form,
**Create draft** stays grey until every required field is filled, and you need a number that can place outbound
calls (see [Phone numbers](/phone-numbers)). See [Campaigns](/campaigns).
Your store has not confirmed its calling consent. An owner clicks **Review and confirm** and confirms under
**Settings**, **Calling consent**. The notice cannot be dismissed: it disappears once you confirm. See
[Calling consent and calling hours](/calling-hours).
Adding documents needs an owner and an agent at **Full control**. Your main agent is often below it: send us the
document and we add it. See [Knowledge](/knowledge).
Ask them to check their junk folder. If it is still missing, remove the invitation and invite them again: **Renew
Invitation** does not send a new email. They must accept with the same email you invited. See
[Members](/members).
## Next steps
Read and listen to every conversation.
Voice, timing, call limits and the rest.
Bring in your customer list, with or without consent.
Call, or message, a group of customers.