Agent Settings
Configure exactly how your AI agent behaves, learns, and communicates.
Personality & Tone
Give your agent a clear name and select its communication style: Professional, Friendly, Concise, or Empathetic. This shapes the stylistic tone of every reply generated.
System Prompting
An optional free-text field for detailed instructions across your whole workspace. Use it to set unbreakable rules like "never discuss competitor products" or "always refer users to the refund policy page". The agent adheres literally.
Confidence Thresholds
Set a strict percentage limit (0–100%, default 70%). If the agent's internal confidence parameter falls below this threshold during generation, the conversation is aborted and instantly escalated to a human agent instead of sending a potentially wrong reply.
Response Length Control
Decide how verbose the agent should be, from tight one-liners for chat widgets to detailed walkthroughs for email. Length is tuned per agent, so each channel gets an appropriate style.
Escalation Settings
When an agent escalates, you orchestrate the handover natively. Enable Auto-assign in Zendesk, trigger a Slack Input Webhook, or send notification emails directly to your QA team.
Test Console & Inspector
Chat with any agent in the built-in test console before it goes live. Simulate a specific customer by ID, read the confidence badge on every reply, and open the View Details inspector to see the full AI context behind each answer: knowledge retrieved, CRM data fetched, and skills called.