Assistant persona
By the end of this page you'll have the widget greeting visitors with your brand's voice and nudging them toward the questions it can answer best.
Before you start
- A bot you've already created (see Create your first bot).
- A rough idea of three to six questions your visitors most often ask.
The persona fields
| Field | Type | Default | Description |
|---|---|---|---|
assistantName | string (min length 1) | Assistant | Displayed in the widget header and in the "typing" indicator. |
welcomeMessage | string | Hi! How can I help? | First message visitors see when they open the widget. |
suggestedQuestions | string[] (max 6) | [] | Clickable prompts rendered below the welcome message. Capped at six entries. |
privacyPolicyUrl | string (URL) or null | null | Link in the widget footer so visitors know how their chats are handled. |
Writing a welcome message that converts
- Lead with what the bot can do, not what it is: "Ask me anything about pricing, integrations, or onboarding" beats "I'm your AI assistant".
- Keep it under two sentences. Visitors who need to read a paragraph will close the widget first.
- Mention a privacy expectation if your audience cares — pair it with a
privacyPolicyUrlfor the one-click follow-through.
Suggested questions
suggestedQuestions is an array of up to six strings. The widget renders them as tappable chips below the welcome message and sends the literal string as the first user turn when tapped.
{
"assistantName": "Acme Bot",
"welcomeMessage": "Ask me about pricing, onboarding, or how Acme compares to alternatives.",
"suggestedQuestions": [
"How much does Acme cost?",
"How long does onboarding take?",
"Does Acme integrate with Slack?",
"What's your refund policy?"
],
"privacyPolicyUrl": "https://acme.example/privacy"
}
Four or five focused chips outperform six generic ones. Watch the Unanswered Questions report during your first week and promote the winners into this list.
A note on the system prompt
The underlying LLM system prompt — the instructions that control tone, refusal behavior, and format — is not yet user-configurable. The bots.system_prompt column exists in the database, but no PATCH endpoint writes to it yet. For now, persona tuning happens through the four fields above. User-editable system prompts are on the roadmap.
Setting persona on bot creation
curl -X POST https://api.mimicbot.app/api/bots \
-H "Authorization: Bearer $MIMICBOT_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"name": "Acme Support Bot",
"clientUrl": "https://acme.example",
"widgetConfig": {
"assistantName": "Acme Bot",
"welcomeMessage": "Ask me about pricing, onboarding, or integrations.",
"suggestedQuestions": [
"How much does Acme cost?",
"How long does onboarding take?",
"Does Acme integrate with Slack?"
],
"privacyPolicyUrl": "https://acme.example/privacy"
}
}'
To edit persona fields after creation, use Settings → Persona in the dashboard.