Playground

The playground lets you test an agent in the dashboard before exposing it to customers.

What to test

• Knowledge retrieval and source grounding.
• Model and instruction changes.
• Lead collection and ticket creation flows.
• Tool or action behavior.
• Escalation and refusal behavior.

Action simulation

When built-in lead and native ticket tools are enabled on your deployment, the playground lets you exercise these actions directly in the chat so you can confirm the agent asks for the right information before it goes live.

These flows are not a dry run: submitting them in the playground creates a real contact or a real ticket in your team's data, the same as a live conversation would. There is currently no simulation mode that previews the action without the side effect, so treat playground testing of these actions as you would testing in a live environment, and clean up any test contacts or tickets afterward.

Deployment readiness

An agent is ready to deploy when it answers common questions correctly, handles unknown questions safely, and performs actions only when the user intent is clear.