Your Project's MCP Server

Every ReadMe project can have its own MCP server. Once enabled, your users' AI coding assistants can connect to it and instantly understand your entire API — endpoints, schemas, authentication, and documentation — without any manual copy-pasting.

Enabling Your MCP Server

1. Open your project in ReadMe and enter Edit Mode
2. Click ✨ AI in the top-right corner to open the AI panel
3. Select MCP and toggle MCP Server on

Your MCP server is now live. Share the URL with your users.

Your MCP Server URL

Standard projects

https://your-project.readme.io/mcp

Enterprise: custom domain

If your project uses a custom domain, your MCP server is at the root of that domain:

https://your-custom-domain.com/mcp

Enterprise: all projects merged

Enterprise accounts also get a root MCP server — a single endpoint that merges all MCP-enabled projects in your organization. This is useful when users work across multiple products or services.

https://your-enterprise.readme.io/mcp

Enterprise: filter to one project

If you only want to expose a specific project from your root domain, use the project filter:

https://your-enterprise.readme.io/mcp?project=project-slug

This limits the MCP server to tools and content from that single project only.

URL Parameters

Append these to any MCP URL to customize behavior.

You can combine parameters:

https://your-enterprise.readme.io/mcp?project=my-api&branch=v2.0

Authentication

Different use cases require different authentication headers. Here's a clear breakdown:

Public projects — no auth needed

If your project is publicly accessible, no headers are required. Users connect with just the URL.

Private projects — control who can read your docs

If your project requires a login to view, users must pass an x-readme-auth header.

Example config for a password-protected project:

{
  "mcpServers": {
    "My API": {
      "url": "https://your-project.readme.io/mcp",
      "headers": {
        "x-readme-auth": "your-site-password"
      }
    }
  }
}

Example config for a teammates-only project:

{
  "mcpServers": {
    "My API": {
      "url": "https://your-project.readme.io/mcp",
      "headers": {
        "x-readme-auth": "bearer rdme_xxxxxxxxxxxx"
      }
    }
  }
}

API execution — authenticate actual API calls

The execute-request tool makes live API calls on your user's behalf. Any headers your users include in their MCP config are automatically forwarded to those API calls.

If your API requires an Authorization header:

{
  "mcpServers": {
    "My API": {
      "url": "https://your-project.readme.io/mcp",
      "headers": {
        "authorization": "Bearer user-api-key-here"
      }
    }
  }
}

Both a private-project header and an API execution header can be combined:

{
  "mcpServers": {
    "My API": {
      "url": "https://your-project.readme.io/mcp",
      "headers": {
        "x-readme-auth": "bearer rdme_xxx",
        "authorization": "Bearer user-api-key-here"
      }
    }
  }
}

Available Tools

Your MCP server exposes two categories of tools.

OpenAPI Tools

These tools are always available and work directly from your OpenAPI specification.

Documentation Tools

These tools require the AI Booster Pack add-on.

📘

Documentation Tools require the AI Booster Pack.

• Startup & Business: upgrade from your Settings → Manage Plan page
• Enterprise: contact your Customer Success Manager

Custom Tools

You can define your own tools — plain-language "functions" that your users' AI assistants can invoke for repeated workflows.

Examples:

• A pre-flight checklist before calling a sensitive endpoint
• A validation reminder that checks required headers are set
• A shortcut that combines multiple endpoint calls into a named operation

To add a custom tool, go to AI → MCP → Custom Tools in your dashboard and click New Tool.

Each custom tool requires:

• Title — unique, letters/numbers/underscores/hyphens/dots only
• Description — when should the AI use this tool?
• Body — the instructions the AI should follow

Tool titles can't match any of the built-in tool names listed above.

Controlling Which Endpoints Are Exposed

By default, all endpoints in your OpenAPI spec are available through MCP. You can disable specific routes you don't want AI tools to access.

Go to AI → MCP → Enabled MCP Routes in your dashboard to toggle individual endpoints on or off.

Generating Setup Instructions for Your Users

Once your MCP server is enabled, click Generate MCP Template in the AI panel. This creates a ready-to-publish doc in your Guides that shows your users how to connect their AI tools — pre-filled with your project's MCP URL and step-by-step instructions for Cursor, Claude Desktop, VS Code, and more.

You can find and edit this doc in a category called MCP SERVER at the bottom of your Guides.

FAQ