Skip to main content

AI Agents

An agent does things a chat reply cannot: it collects a lead's details and scores them, opens a support ticket with a full conversation summary, or checks a calendar and books a meeting — all without the visitor leaving the chat.

Available Agents

AgentWhat it doesPlanStatus
Lead QualificationSpots buying intent, collects BANT details, scores the lead 1–10Free and aboveBeta
Technical SupportAnswers from your knowledge base, escalates as a structured ticketFree and aboveBeta
Meeting SchedulingChecks real calendar availability and books the meetingStarter and aboveBeta

All three are in beta: they work in production, and their behaviour is still being refined.

Enabling an Agent

  1. Open your agent in the dashboard
  2. Go to AI Agents
  3. Select the agent you want
  4. Toggle it on
  5. Work through its readiness checklist

Agents are configured per chatbot, so one chatbot can qualify leads while another only answers support questions.

What Agents Cost

Nothing beyond the conversation itself. Agent actions are included on every plan and are not metered separately — a reply that books a meeting costs the same single message credit as a reply that answers a question. See Quotas.

The plan tier decides which agents you may run, not how many actions you may take.

Readiness Checklists

Each agent shows a checklist before it can do useful work. Typical items:

  • Agent enabled — the toggle itself
  • Manual review — a setting you should confirm rather than accept blindly, such as scoring thresholds
  • Calendar integration — a connected Google or Microsoft account (Meeting Scheduling only)
  • Calendar assignment — which calendars this chatbot may book into (Meeting Scheduling only)

An agent with an incomplete checklist stays inactive rather than acting on bad configuration.

Agents and Your Knowledge Base

Agents work on top of the same knowledge base your chatbot answers from. A well-trained chatbot makes a better agent: the Technical Support agent escalates less when your documentation is complete, and the Lead Qualification agent asks smarter follow-up questions when it can already answer product questions.

See Training Overview before enabling agents.

Testing Agents

Use the Playground to test agent behaviour. Playground conversations do not consume message credits, and playground actions are tracked separately from production activity.

:::warning Real Side Effects Some agent tools have real effects — the Meeting Scheduling agent creates real calendar events. Test with a calendar you don't mind writing to. :::

Next Steps