> ## Documentation Index
> Fetch the complete documentation index at: https://docs.a2v2.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Sandbox

> Chat with your agent, tune its model and instructions, and watch how it answers — all in one place.

The **Sandbox** is your agent's control room and test chat in one. As its subtitle
says, it's where you "adjust your agent's settings and preview responses in real
time" — set the model, temperature, instructions, and chat options, then talk to
your agent exactly the way a visitor would, with a live preview beside the
settings.

Open **Agent → Sandbox** from the agent sidebar.

## Prerequisites

* An agent (see [Create an agent](/building/create-agent)).
* At least one [Completed knowledge source](/knowledge-base/overview) for the
  agent to answer from.

## What's on the Sandbox

| Control                        | What it does                                                                                                                                                |
| ------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Choose Your AI Model**       | Pick the AI model that powers answers. Each option shows its credits/msg and any HIPAA/Premium badges. See [Model & temperature](/building/model-settings). |
| **Temperature**                | Slider for how focused vs. creative answers are, from **Reserved** to **Creative** (default 0.5).                                                           |
| **Allow file uploads in chat** | Lets visitors attach files in the chat. Supports PDFs, documents, and images, up to 5MB per file.                                                           |
| **Allow voice input in chat**  | Lets visitors record voice messages (up to 30 seconds). On by default.                                                                                      |
| **Instructions**               | Shows the active instruction version (for example, **V1.0 · Active**) with a shortcut to **Manage instructions**.                                           |
| **Live preview**               | A working chat with your agent for instant testing.                                                                                                         |
| **Undo / Redo**                | Step backward or forward through your unsaved setting changes.                                                                                              |
| **Save Changes**               | Applies your settings to the agent's live answers.                                                                                                          |

<Frame>
  <img src="https://mintcdn.com/a2v2ai/ItYQQYTHsiiRmr-t/images/testing/sandbox.png?fit=max&auto=format&n=ItYQQYTHsiiRmr-t&q=85&s=df8e36ed37cdb98119ee277eab14888e" alt="The Sandbox with the settings panel and live chat preview." width="1440" height="900" data-path="images/testing/sandbox.png" />
</Frame>

<Note>
  Changes you make in the Sandbox apply to your live agent once you select **Save
  Changes**. Use **Undo** and **Redo** to step through edits before saving.
</Note>

## Test your agent

<Steps>
  <Step title="Ask a real question">
    Type a question your knowledge base should answer.
  </Step>

  <Step title="Watch the answer stream">
    The agent replies in real time. While it works, you may see it *thinking* and
    using tools before the final answer appears.
  </Step>

  <Step title="Check the sources">
    Confirm the answer cites the right source from your knowledge base — that's how
    you know it's grounded in your content.
  </Step>

  <Step title="Adjust and retry">
    Tweak the model, temperature, or [instructions](/building/instructions), then
    ask again to compare.
  </Step>
</Steps>

<Note>
  Testing in the Sandbox is for your own iteration. Use it freely to dial in the
  agent before you [embed it](/embed/install).
</Note>

## Tips

<Tip>
  * Test the **hard questions** — edge cases, ambiguous phrasing, and things just
    outside your content — to see how the agent handles them.
  * If an answer is wrong, check whether the source is **Completed**, then sharpen
    your [instructions](/building/instructions) or add a [Q\&A](/knowledge-base/qa).
  * To compare two full configurations at once, use the
    [Playground](/testing/playground).
</Tip>

## Troubleshooting

<AccordionGroup>
  <Accordion title="The agent says it doesn't know">
    Make sure a relevant source shows **Completed** in the
    [Knowledge Base](/knowledge-base/overview). If it does, add a closer-matching
    [Q\&A pair](/knowledge-base/qa).
  </Accordion>

  <Accordion title="Answers are off-tone or off-topic">
    Refine your [instructions](/building/instructions) and consider lowering
    [temperature](/building/model-settings#tune-temperature).
  </Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Model & temperature" icon="microchip" href="/building/model-settings" />

  <Card title="Instructions" icon="pen" href="/building/instructions" />

  <Card title="Playground" icon="code-compare" href="/testing/playground" />

  <Card title="Embed your agent" icon="code" href="/embed/install" />
</CardGroup>