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

# Test your agent

> Learn how to test and fine-tune your AI Agent using the Playground to improve accuracy before going live.

The **Playground** lets you simulate real conversations so you can evaluate how your AI Agent understands, responds, and applies the knowledge you’ve provided.

**Testing** before going live helps you spot issues early and **fine-tune** your AI Agent’s tone, accuracy, and use of sources.

## Using the Playground

<Steps>
  <Step title="Open your AI Agent">
    Navigate to **Agents**, then select the agent you would like to test.

    <Frame>
      <img src="https://mintlify.s3.us-west-1.amazonaws.com/watermelon/images/features/website/website-source-select-agent.png" alt="Website Source Select Agent Pn" />
    </Frame>
  </Step>

  <Step title="Open the Playground">
    Within the Agent menu, select the **Playground**. Here you’ll find useful tips, functionalities, and limitations to help you get the most out of your testing.

    <Frame>
      <img src="https://mintcdn.com/watermelon/0CHjAwnwcivzIXmQ/images/get-started-quickstart-playground-1.png?fit=max&auto=format&n=0CHjAwnwcivzIXmQ&q=85&s=b9ee216c54b819ecc441d12957724a2c" alt="Get Started Quickstart Playground 1" width="2880" height="1460" data-path="images/get-started-quickstart-playground-1.png" />
    </Frame>
  </Step>

  <Step title="Start Testing">
    Now you can send messages to your AI Agent, as a customer would.

    1. Try to imagine what questions your **customers** might ask your AI Agent.
    2. Type a question and press Enter to see how your AI Agent responds based on its Domain Knowledge and Instructions.
    3. Below each response, you’ll see which **sources** the Agent used to generate its answer.
    4. Ideally, test your agent with **50-100** frequently asked end-user questions before going live.

    This helps you understand how your AI Agent interprets your knowledge base and whether it’s applying the right information.

    <Note>
      **Conversation forwarding** does not function inside the Playground. You can still test how your AI Agent reacts to triggers, but conversations aren’t actually sent to the Inbox until the Agent is live.
    </Note>
  </Step>

  <Step title="Provide Feedback">
    Next to each AI Agent response, you’ll see **thumbs-up** and **thumbs-down** icons.

    • <Icon icon="thumbs-up" /> Click the thumbs-up if the answer is correct.

    • <Icon icon="thumbs-down" /> Click the thumbs-down if it needs improvement.

    <Frame>
      <img src="https://mintcdn.com/watermelon/Dbi8JW4Z7cKWYgN1/images/add-sources-thumbs-down.png?fit=max&auto=format&n=Dbi8JW4Z7cKWYgN1&q=85&s=9ab353ddc440b86cf990db932eca60eb" alt="Add Sources Thumbs Down" width="2880" height="1460" data-path="images/add-sources-thumbs-down.png" />
    </Frame>

    When you select the thumbs-down, a **pop-up** appears where you can:

    1. Choose the **category** for the new instruction (or create one).
    2. Add information explaining how to **improve the response** next time.

    <Note>
      For best results, write this as an *instruction*, not the literal response you want to see.
    </Note>

    All feedback and instructions added here appear under the **Instructions** section of your AI Agent.

    <Note>
      The instruction given **will not** carry the context from the test conversation, so it is important to provide it, for example:

      \
      *<u>"If the customer asks about returns, mention that returns are possible at their nearest Company store as well as online."</u>*
    </Note>

    <Frame>
      <img src="https://mintcdn.com/watermelon/Dbi8JW4Z7cKWYgN1/images/add-sources-playground-instruction.png?fit=max&auto=format&n=Dbi8JW4Z7cKWYgN1&q=85&s=6d9b6e12e82f362eabf4182528ebe4fe" alt="Add Sources Playground Instruction" width="2872" height="1462" data-path="images/add-sources-playground-instruction.png" />
    </Frame>
  </Step>
</Steps>