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

# MCP Server

> Serve your MCP Apps to AI hosts like ChatGPT.

<img src="https://cdn.sunpeak.ai/chatgpt-counter.png" alt="MCP App in ChatGPT" />

## Overview

The [MCP (Model Context Protocol)](/mcp-apps/mcp/overview) server lets you serve your apps to AI hosts like ChatGPT and Claude. Once running, the host can discover and render your MCP Apps.

During development, `sunpeak dev` runs an MCP server that serves simulation fixture data. For production, use `sunpeak build` followed by `sunpeak start` to run your real tool handlers — see the [Deployment Guide](/app-framework/guides/deployment).

## What is MCP?

MCP is the protocol AI hosts use to communicate with your servers. It handles [tool](/mcp-apps/mcp/tools) and [resource](/mcp-apps/mcp/resources) registration, discovery, and invocation.

<Info>
  Learn more in the [MCP primer](/mcp-apps/mcp/overview) or the [Model Context Protocol documentation](https://modelcontextprotocol.io/).
</Info>

For details on how MCP Apps extends the protocol with interactive UIs, see the [MCP Apps documentation](/mcp-apps/introduction).

## Quick Start

<Tabs>
  <Tab title="Development">
    Start the development server:

    <Tabs>
      <Tab title="pnpm">
        ```bash theme={null}
        pnpm dev
        ```
      </Tab>

      <Tab title="npm">
        ```bash theme={null}
        npm run dev
        ```
      </Tab>

      <Tab title="yarn">
        ```bash theme={null}
        yarn dev
        ```
      </Tab>
    </Tabs>

    This starts both the inspector and an MCP server with Hot Module Reload.
    The MCP server runs on port `8000` and serves simulation fixture data.
  </Tab>

  <Tab title="Production">
    Build and start the production server:

    <Tabs>
      <Tab title="pnpm">
        ```bash theme={null}
        pnpm build && pnpm start
        ```
      </Tab>

      <Tab title="npm">
        ```bash theme={null}
        npm run build && npm run start
        ```
      </Tab>

      <Tab title="yarn">
        ```bash theme={null}
        yarn build && yarn start
        ```
      </Tab>
    </Tabs>

    This compiles your tools, resources, and optional auth, then starts a production MCP server with real handlers.
  </Tab>
</Tabs>

Then connect to a host using a tunnel. See [runMCPServer API Reference](/app-framework/run-mcp-server#connecting-to-chatgpt) for the full connection flow.

## Dive Deeper

<Card horizontal title="runMCPServer API Reference" icon="code" href="/app-framework/run-mcp-server">
  Complete configuration options and host connection guide.
</Card>

<Card horizontal title="Deployment Guide" icon="rocket" href="/app-framework/guides/deployment">
  Build, start, and deploy your app to production.
</Card>

<Card horizontal title="Simulation API Reference" icon="code" href="/testing/simulations">
  Complete simulation documentation.
</Card>