Configure multilingual agent prompts and hangup settings

The Agent Tab controls your agent’s welcome message, prompts, and conversation behavior. Add multiple languages, set a default, and write a dedicated prompt for each.

Agent welcome message

The first thing callers hear when they connect.

Keep it short

Brief greetings work best. Long announcements feel robotic.

Use variables

Personalize with {{variable_name}}, e.g. {{customer_name}}.

Welcome message behavior

Setting	What it controls
Ignore user speech before welcome message	Discards speech received before the welcome message finishes playing.
Welcome Message Delay (in ms)	Waits `0–1000` milliseconds before the agent starts the welcome message.

Welcome message per language

When the agent has two or more languages, a Translations button appears inside the Agent Welcome Message field. Click it to open the Translations panel, which has one box per language:

Translations side panel with a Hindi row marked Default holding the agent welcome message, an English row with its own welcome message and a Translate icon, and Save changes and Discard buttons at the bottom — Translations panel with the default Hindi welcome message and an English welcome message

Default language: listed first and marked Default. Its box is the Agent Welcome Message itself, so editing it here edits the field above.
Other languages: optional. Leave a box empty and calls in that language play the default welcome message.
Translate: the icon in a language’s box fills it with a translation of the default welcome message. Review or edit the result before saving. The icon shows once the agent is saved and the default message is not empty.

Click Save changes to apply the panel to the agent, then Save agent to save it. Discard closes the panel without applying your edits. Each call opens with the welcome message of the language it starts in. Per-language messages support the same {{variable_name}} tokens, and their variables show up with the rest for test calls. To start a call in a particular language, see Start a call in a chosen language.

Canvas

This section covers the canvas where each language has its own prompt. New agents in accounts with the Single Prompt view (Beta) use one shared prompt for all languages instead. See Single Prompt view to check which view your agent uses.

Each language gets its own prompt. Select a language tab and write the prompt for that language. The agent activates the matching prompt when speaking in that language during a call.

Canvas prompt editor with Hindi (Default) and English language tabs, Add Language, AI Edit and Variables buttons, a prompt with highlighted candidate_name, city, role and screening.score variables, Add Section and a token count — Canvas with Hindi as the default language, an English tab, highlighted variables, the Variables button and the token counter

Managing languages

Languages are added and removed here, in the Agent Tab. The Languages Tab shows the same set of languages and is where each one gets its own voice and transcription providers.

Add a Language

Click + Add Language to create a new language tab.

Set the Default

Click the crown icon next to any language to make it the default. The default language is what the agent starts every conversation in, marked with (Default) in the tab.

Write Per-Language Prompts

Select each language tab and write its prompt in that language’s native script — see Non-English Prompts for guidelines. Don’t include language-switching instructions in it; switching is automatic — see How Language Switching Works.

Remove a Language

Click the x on a tab to remove it. The default language cannot be removed — promote another language first.

Make English default tooltip explaining that the crown icon sets the default language for the main prompt and multilingual settings — Crown icon tooltip to set a language as the default

Prompt structure

Recommended Prompt Sections

Section	Purpose	Example
Personality	Tone and feel	”warm, perceptive, and results-driven”
Context	Role background	”You are calling on behalf of Acme Corp…”
Instructions	Tasks and flow	”Ask for their order number first…”
Guardrails	Restrictions	”Never discuss competitor products…”

The editor shows a token count in the bottom-right to help you stay within LLM limits.

Variable syntax

Syntax	Purpose	How to use
`{{variable_name}}`	Insert or define variables	Type `{` to open the variable picker. Select an existing variable or type a new name to create one.
`{variable_name}`	The same variable, single-brace form	Still accepted, e.g. for prompts carried over from an earlier agent. The editor keeps whichever form you typed.
`{variable.path}`	Read into a variable that holds a JSON value	Dot and bracket notation both work — `{prior.score}`, `{prior[loans][0][amount]}`. Only the root (`prior`) is a variable you fill in. See Adding Variables.
`@`	Insert prompt modules, custom functions, or variables	Type `@` to browse and select existing modules, functions, or variables. You cannot create new items with `@`.

Variables with {{ }}

Select existing variables or create new ones by typing a name. Values are passed via API or CSV at call time.

Modules and functions with @

Select existing prompt modules, custom functions, or variables. Cannot create new items.

Browse modules

Click the Browse Modules button in the top-right of the prompt section to open the full modules library. Browse by category (Collection, Optional, Flow, Sector, Universal), preview what each module does, and insert it directly into your prompt. See the Prompting Guide for the full list of available modules.

AI edit

Click AI Edit in the top-right of the prompt section to open a chat panel and rewrite the prompt by describing the change you want. The rest of the Agent Tab steps aside while the panel is open, so the prompt gets the full height of the screen.

AI Edit is in beta and is free to use during the beta.

Ask for a Change

Describe the change in the composer, e.g. “ask for the order number before anything else” or “make the tone warmer and cut the closing script down”. Each message is answered against whatever the prompt says at that moment, so you can keep editing in the Canvas between messages.

Review the Proposed Changes

A rewrite comes back as a diff — added lines in green, removed lines in red. Click the expand icon on the Proposed changes card to read a long rewrite at full height.

Approve or Discard

Approve edit writes the rewrite into the Canvas and saves the prompt. Discard leaves the prompt untouched. Every proposal has to be approved or discarded before you can send the next message.

The session works on one language’s prompt. The language you had selected when you opened the panel is shown as a read-only label, and languages cannot be switched or added until you close the panel. Reopening the panel for the same agent and language picks the conversation back up.

Approving an edit saves the prompt, but not the agent. Click Save agent for the new prompt to be used on calls.

Revert AI edit change

While the panel is open, Revert AI Edit change in the top-right puts the prompt back to how it was before the editing session started. It becomes available once you have approved at least one edit, and asks for confirmation because everything you approved during the session is discarded.

The revert point only lasts as long as the session — closing the AI Edit panel discards it. Approved edits are already saved at that point and stay in the prompt.

Per-language hand-in settings

Agent name and Hand-in message — both optional — are set per language in the Languages Tab: how the agent introduces itself in a language and what it says while switching into it.

Variables panel

Every {{variable_name}} in your prompt or welcome message automatically appears in the Variables panel. Click Variables at the top right of the Canvas to open it, then fill in test values to preview how the prompt behaves before going live. The values are sent as call context (user_data) when you place a test call.

Timezone: the selector in the panel header sets the timezone context for test calls. It applies to all languages.
Language tabs: when the agent has two or more languages, each tab lists the variables used in that language’s prompt plus those in the welcome message. The crown marks the default language. A value is shared across languages, so a variable used in several prompts only needs to be filled once.
Long values: hover over a field and click the expand icon to edit a long or multi-line value in a larger box.
Filter: with more than six variables, a filter box appears at the top to find a variable by name.
The footer shows how many variables the agent has and how many are still empty.

The panel closes while AI Edit is open.

JSON variables

A variable the prompt reads into with a path, like {screening.score}, holds a whole JSON value rather than a single word. It gets a taller JSON editor with syntax highlighting, an error marker on the offending character, and a Format button that re-indents the value. A value that is not valid JSON yet is flagged, because path references cannot resolve against it.

Import and export values

Use Import and Export in the panel footer to fill in or save test values in one go.

Export menu in the Variables panel footer with Export JSON (all languages, current values) and Export CSV (batch upload format, one column per variable) — Export options in the Variables panel footer

Option	What it does
Export JSON	Downloads `<agent-name>-variables.json`: one object with every variable, its current value and the `timezone`.
Export CSV	Downloads `<agent-name>-variables.csv` in the batch calling upload format: one column per variable, one row of values.
Import JSON	Upload a `.json` file or paste JSON in the same shape as the export, e.g. `{ "candidate_name": "Riya Sharma", "timezone": "Asia/Kolkata" }`.
Import CSV	Upload a `.csv` file with a header row of variable names and a row of values.

Import JSON dialog with Upload file and Paste tabs, a JSON example with candidate_name, role, city, screening and name, a Sample template card with a Download link, and Cancel and Import buttons — Import JSON dialog with the Paste tab and a sample template to download

Each import dialog has a Sample template you can download, prefilled with the agent’s variables and current values. Imported fields are briefly highlighted. Keys that aren’t variables of this agent are skipped, and if none match, nothing is imported.

Hangup using prompt

Let your agent decide when to end calls based on conversation context instead of silence detection or timeouts.

Hangup using a prompt card with the toggle switched on and a prompt describing what to do when the booking is confirmed, the user is not interested, busy, or it is a wrong number — Hangup using a prompt enabled, with the completion conditions the agent hangs up on

Enable the Toggle

Turn on Hangup using a prompt.

Define Completion Conditions

Write conditions for when a conversation is complete. For multilingual agents, include closing lines in each language.

Example: Multilingual Hangup Prompt

A conversation is considered complete if any of the following conditions are met:
User is not interested:
The assistant has said the following closing line:
  Conversation Closing (English): "That is sad to hear. But no worries if you would ever want to learn more give me a cool."
  Conversation Closing (Hindi): "ये सुनकर अफ़सोस हुआ। लेकिन कोई बात नहीं, अगर आप कभी और जानना चाहें तो मुझे call कर लें।"
The user has responded after that with any of the following phrases: "goodbye", "bye", "thank you", "thanks", "ok", "okay", or similar.
The last message in the transcript must be from the user.

Without this, calls end only on silence detection or timeouts.

Next steps

Intelligence tab

Configure the language model and knowledge bases

Languages tab

Set up voice, transcription, and per-language hand-in

Prompting guide

Best practices for writing prompts

Using variables

Dynamic personalization with context

How language switching works

What actually triggers a mid-call switch

Build

Deploy

Monitor

Go live

Configure multilingual agent prompts and hangup settings

Agent welcome message

Keep it short