Testing agents with chat

Test your agents directly in the dashboard using the built-in chat interface before deploying them to production. This lets you iterate on configuration without writing any integration code.

Open the chat

Navigate to AI Agents in the dashboard sidebar and click on an agent to open its detail page. The chat panel is available on the right side of the agent detail page.

How it works

Type a message in the chat input and press Send. The agent responds using its configured model, system prompt, tools, and knowledge bases. You see the full response as it generates, giving you immediate feedback on how the agent behaves with your current settings.

Streaming responses

Agent responses are delivered via Server-Sent Events (SSE), appearing progressively as the model generates them. This means you do not need to wait for the entire response to complete before reading the output.

Session management

Chat sessions are ephemeral and used for testing only. Each session maintains conversation history for the duration of your browser session, allowing multi-turn conversations. Sessions are not persisted between browser sessions.

Note

Test chat sessions are separate from production API conversations. They do not appear in logs or analytics.

Tool execution

When the agent uses tools during a conversation, the chat interface displays the tool calls and their results inline. This gives you visibility into:

• Which tools the agent chose to call
• The parameters the agent passed to each tool
• The results returned by each tool
• How the agent incorporated tool results into its response

This is especially useful for debugging tool selection and verifying that custom tools return the expected data.

Iterating on configuration

Tip

Use the test chat to iterate on your system prompt and tool selection. Adjust settings, save the agent, and test again until you get the behaviour you want.

A typical iteration workflow:

1. Open the agent detail page and start a test chat
2. Send messages that represent real use cases
3. Review the agent’s responses and tool usage
4. Adjust the system prompt, temperature, tools, or skills as needed
5. Click Save to apply changes
6. Test again with the same or similar messages

Next steps

• Agent configuration — adjust settings based on your test results
• Tools — create custom tools if the agent needs additional capabilities
• Skills — add reusable instruction sets to refine agent behaviour