Google Workspace MCP Server (Railway Deployment)

A robust, generic Model Context Protocol (MCP) Server that integrates AI agents with Google Workspace. It acts as a bridge, enabling any standard MCP-compliant AI agent (such as Claude Desktop) to securely interact with Gmail and Google Docs.

This version has been optimized for deployment to cloud providers like Railway using an Express-based HTTP server with Server-Sent Events (SSE) and API Key authentication.

Provided Tools

1. gmail_draft_email: Creates a draft email in your Gmail account without sending it.
2. gmail_send_email: Sends an email directly from your Gmail account.
3. docs_append_content: Appends text content to the end of a specified Google Document.

Setup Instructions

1. Configure Environment Variables

This server expects the following environment variables to securely connect to Google APIs without relying on local files:

• PORT: (Optional) Port for the Express server (default: 3000). Provided automatically by Railway.
• MCP_API_KEY: A secure random string you choose to protect your endpoint (e.g., sk-mcp-12345...).
• GOOGLE_CLIENT_ID: Your Google OAuth 2.0 Client ID.
• GOOGLE_CLIENT_SECRET: Your Google OAuth 2.0 Client Secret.
• GOOGLE_REFRESH_TOKEN: The refresh token generated after your initial OAuth consent flow.

2. Deploy to Railway

1. Push this repository to GitHub.
2. Create a new project in Railway and select the repository.
3. Railway will automatically build the TypeScript project via npm run build and start it.
4. Add the required environment variables in the Railway dashboard.
5. Generate a public domain for your Railway service (e.g., https://my-mcp-server.up.railway.app).

3. Configure Your MCP Client

Instead of running a local node process, you must configure your AI agent to connect via SSE over HTTP. Provide the API key in the environment to authenticate.

Claude Desktop Configuration

Add this server to your claude_desktop_config.json:

{
  "mcpServers": {
    "google-workspace": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/inspector", "http://my-mcp-server.up.railway.app/sse"],
      "env": {
        "MCP_API_KEY": "your_secure_api_key_here"
      }
    }
  }
}

(Note: To connect Claude Desktop directly to a remote SSE server, you typically use a specialized SSE client wrapper, or pass the authorization header via the client's built-in HTTP connector if supported).