> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xpressbot.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Test

> Try the agent yourself, with real messages, before a customer ever sees it — the Agent Playground.

## What is this?

The **Test** tab is the **Agent Playground** — a private chat window where you talk to your own agent exactly like a customer would, before its Status goes Active.

\[SCREENSHOT: ai-agent-test-playground — The Test tab's Agent Playground with a sample conversation]

## What can I do here?

* Type a message and press **Send** to see the agent's real reply.
* See which tools or capabilities the agent used to answer.
* Click **New Conversation** to reset and start a fresh test session.

## How to use it

### Test every time you change Instructions or Knowledge

This is the step people skip, and the one that catches problems before a real customer does. Any time you edit [Instructions](/workspace/ai-agents/tabs/instructions) or add a [Knowledge](/workspace/ai-agents/tabs/knowledge) source, come back here and re-test — don't assume a change worked just because it reads correctly.

### Try to break it, not just confirm it works

Test the golden path, but also:

* Something it shouldn't know the answer to.
* Something rude or off-topic.
* Something outside its stated job, to see if it hands off correctly.
* If it has [Capabilities](/workspace/ai-agents/tabs/capabilities) enabled, a message that should trigger one — confirm it actually fires and that the reply makes sense.

### Start a New Conversation between unrelated tests

The Playground remembers context within one session the same way a real conversation would. Reset with **New Conversation** before testing an unrelated scenario, or an earlier exchange can influence a later reply in ways a fresh customer conversation never would.

## Troubleshooting / Technical Notes

* **An answer is wrong.** Fix [Instructions](/workspace/ai-agents/tabs/instructions) or add the missing fact to [Knowledge](/workspace/ai-agents/tabs/knowledge), then test again — don't guess which one fixed it without re-testing.
* **A capability didn't fire when you expected it to.** Check the capability (or the triggered automation's Description) rather than assuming the Playground itself is broken — see [Capabilities](/workspace/ai-agents/tabs/capabilities).

## Related docs

* [Instructions](/workspace/ai-agents/tabs/instructions)
* [Knowledge](/workspace/ai-agents/tabs/knowledge)
* [Behavior](/workspace/ai-agents/tabs/behavior) — set Status to Active once testing looks right