Testing your AI in the Playground

The Playground simulates a real customer conversation with your AI assistant, using your actual configuration — the same Knowledge Base, catalog, tools, and guardrails a real customer's message would hit.

Every reply comes with a trace: which tools it called, which Knowledge Base passages or saved answers it relied on, and why it answered the way it did. Use this to catch a wrong or unbacked answer before it reaches a customer, not after.

If a reply looks off, the fix is almost always in your Knowledge Base, catalog data, or Agent Configuration — not the Playground itself. Adjust the source, then re-test the same question.