Tabs
AI Agent tabs
Main settings
Name, goal, personality, tone, fallback, custom instructions, suggested questions, domains, lead capture, and handoff.
Training
Add, review, retrain, or delete sources and inspect characters, knowledge sections, crawl pages, and errors.
Widget
Customize colors, launcher label, bubble icon, avatar, welcome copy, position, and branding.
Playground
Test real visitor questions before public launch and inspect answer behavior.
Embed
Review readiness checks and copy the install snippet after the bot is ready.
Conversations
Review conversation rows and full transcripts with sources, context, and attachments.
Channels
Connect WhatsApp, Messenger and Instagram. Business email, for answering your company inbox, is coming soon.
Contacts
Manage captured leads and update lead status.
Analytics
Review quality, usage, visitor context, response time, estimated cost, and training gaps.
Behavior
Behavior settings
- Changing lead capture or handoff rebuilds the runtime flow used by playground and widget turns.
- Suggested questions should be real customer questions, not generic prompts.
- Fallback text should clearly say when the assistant does not know enough.
- Custom instructions should describe behavior, not private prompts or secrets.
Domains
Domains
A published AI Agent must have an allowed domain or explicitly allow any domain.
Access
Access
What someone can change on an AI Agent comes from their role: Editor and Admin change settings, knowledge and the widget and can publish, Member moves leads, and Support answers conversations. A member can also be limited to chosen AI Agents, and the others stay hidden from them.
Train AI Agent on website content
Train an AI Agent on approved website pages, documents, media, text, and manual Q&A sources while reviewing status, stale content, and gaps.