Playground Overview

The Playground is a real-time chat interface for testing agents before they go live. Select an agent and application, start a conversation, and see exactly how the agent will behave for your users.

Why use the Playground?

Before deploying an agent to a production widget, you want to make sure it responds correctly, references the right knowledge, and handles edge cases gracefully. The Playground gives you a safe environment to:

• Validate agent behavior — Confirm the agent answers questions accurately and follows your instructions.
• Test instruction changes — Iterate on agent instructions and immediately see the impact.
• Debug issues — Reproduce and diagnose problems without affecting real users.
• Try all interaction modes — Test Tell, Show, and Do modes in a single session.

How it works

1. Navigate to the Playground page.
2. Select an agent and its parent application.
3. Start chatting — messages stream in real time, so you see the agent thinking and responding live.

The Playground uses the same infrastructure as your production widgets. What you see here is what your users will experience.

Interaction modes

The Playground supports all three interaction modes:

Mode	What the agent does
Tell	Responds with text answers based on its knowledge
Show	Demonstrates a workflow by navigating your application while you watch
Do	Takes actions directly in your application on your behalf

Tip

Start with Tell mode to verify knowledge accuracy, then move to Show and Do to test interactive workflows.