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

# Playground

> Test models, protocols, policies, and tool conversations from the dashboard.

Playground lets you try a project before integrating an application. Each model round uses the
selected Aixy API key's project, model access, guardrails, and budgets, so you can verify the same
configuration your client will use.

## Start a conversation

1. Select the project and open **Playground**.
2. Choose a model from the picker.
3. Open **Request settings** and select one of your project API keys.
4. Choose **Protocol / endpoint** for Chat Completions, Responses, or Messages.
5. Send a short message and inspect the streamed reply and response metrics.

Create a key from **API keys** if the project has none. The dashboard uses the key's identity
without displaying its saved secret. Responses in Playground are limited to 1,024 output tokens.
Application requests choose their own supported output bound.

## Check policies and costs

With a hard budget, Playground checks price coverage before the model request. That check sends
no prompt and reserves no spend. A notice identifies incomplete pricing; use **Recheck prices**
after an administrator has [completed the pricing](/guides/pricing#resolve-a-playground-pricing-warning).
Normal budget admission still runs when you send the message.

Inspect the request ID and open [Activity](/observe/activity) to see the provider error, route
attempts, token usage, and available cost evidence. An error on the first model round preserves
your draft so you can correct it and retry deliberately.

## Reproduce a text request

Use the cURL action to copy the configured text request, then supply your key from a protected
local environment. This is useful when moving from a successful dashboard test to a backend or
when comparing client behavior. See the [quickstart](/quickstart) for runnable examples.

## Add tools and manage conversation state

Choose **Select tools** to use your connected accounts. Playground searches for authorized actions
as needed and asks you to approve each external action. Follow
[Use tools in Playground](/agent-tools/playground) for setup and execution behavior.

**New conversation** clears the current history. Changing model, key, protocol, or selected tools
also starts a new conversation. **Stop** cancels the current turn. Conversation state lives in the
current page and is cleared on navigation or reload; explicitly enabled Activity capture follows
its separate [retention policy](/observe/activity#configure-content-capture).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.