> ## Documentation Index
> Fetch the complete documentation index at: https://docs.0mcp.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Playground

> Test your MCP server before connecting an AI client.

The **Playground** lets you interact with your published MCP server through a chat interface. It's the fastest way to verify that your [tools](/capabilities/tools), [resources](/capabilities/resources), and [prompts](/capabilities/prompts) behave as expected before using them in applications like Claude Code or Cursor.

## Before you start

Before using the Playground, make sure you have:

* Published a version of your MCP server
* An AI provider API key, if required
* Any upstream API credentials needed by your tools

## Using the Playground

1. Publish your MCP server.
2. Open **Playground** from your server.
3. Add your AI provider API key if prompted.
4. Provide any required upstream API credentials.
5. Start chatting and invoke your tools, resources, or prompts.
6. Review the execution logs to see how each MCP request was processed.

<Note>
  The Playground always uses the **currently published version** of your MCP server. Draft changes are not available until you [publish a new version](/guides/versioning).
</Note>

## What you can test

The Playground supports everything available in your published server, including:

* Tools
* Resources
* Prompts

This allows you to verify behavior before connecting external MCP clients.

## Best practices

* Start by testing read-only tools.
* Verify tool inputs and outputs before exposing write operations.
* Confirm [authentication](/concepts/authentication-model) is working correctly.
* Review [execution logs](/guides/logs) when troubleshooting unexpected results.

<Warning>
  Test operations that create, update, delete, send, or charge data with a limited test account before using production credentials.
</Warning>

## Privacy

Your credentials stay under your control.

* AI provider API keys are stored only in your browser's session storage.
* Upstream API credentials are used only to execute requests.
* 0mcp does not store Playground credentials, request payloads, or response payloads as activity data.

<Info>
  Activity records keep only the operational details needed for usage and troubleshooting, such as status, response time, capability name, and data size.
</Info>