Summary
The Playground lets you test your web AI agent before deploying it to your website. You can preview how the agent responds, test its instructions and behavior, try voice interactions, and review the web widget experience.
The Playground does not use a real WhatsApp number or conversation. As a result, Wati actions that require an actual conversation cannot be tested in the Playground.
To test Wati actions, such as assigning a conversation to a user, deploy the agent to your Wati account and test it using a real WhatsApp session.
Instructions
Use the Playground to test and refine your AI agent before deploying it. The Playground provides a web-based testing environment that closely matches the experience users will have with your deployed web widget.
Test your AI agent in the Playground
The Playground provides an integrated environment to test your agent's behavior and user experience.
In the Playground, you can:
Test your text agent's behavior and responses
Test voice interactions with your voice agent
Update agent instructions using Astra Vibe to adjust agent behavior
Preview instruction changes in real time
Customize the agent's appearance using available branding options
Review how users will interact with the web widget
This helps you identify and fix issues with your agent's instructions, responses, and web experience before deployment.
Wati actions can't be tested in the Playground
The Playground does not use an actual WhatsApp number or live WhatsApp conversation. Because Wati actions require a real conversation, these actions cannot be fully tested in the Playground.
For example, if your agent is configured to assign a conversation to a user, the assignment cannot happen in the Playground because there is no real WhatsApp conversation to assign.
Test Wati actions after deployment
To test Wati actions:
Deploy the AI agent to your Wati account.
Start a conversation using a real WhatsApp session.
Trigger the condition that runs the Wati action.
Verify that the action works as expected in Wati.
For example, if your agent is configured to assign a conversation to a specific user, use a real WhatsApp session to start a conversation and verify that the conversation is assigned correctly.
Note: Use the Playground to test your agent's responses, instructions, voice interactions, and web widget experience. Use a real WhatsApp session to test Wati actions that require an actual conversation.
If an action cannot be tested in the Playground, you’ll see a message explaining that the action failed to trigger, isn't available in the playground and you need to test it in live mode.
Improved testing experience
The Playground is designed specifically for testing web AI agents and provides a more realistic preview of the deployed web widget.
You can use it to:
Preview the web widget experience
Test different conversation scenarios
Review agent responses
Refine agent instructions
Preview changes before deployment
Consistent visual experience
The Playground includes visual improvements that help provide a consistent experience between testing and deployment.
These improvements include:
Optimized default backgrounds
Improved screenshot fallback handling
Better visual consistency between the Build Demo page and the Playground
As a result, the Playground provides a closer preview of how your web AI agent will appear and behave when deployed to your website.