google -remote

Google MCP Remote exposes a Cloudflare Workers-based MCP server that grants an AI client (like Claude or Cursor) access to Google services such as Gmail, Calendar, Drive, Tasks, YouTube, and Contacts through the MCP interface. After deploying your own instance, configure your AI client to point to the server's /sse endpoint to begin issuing natural language prompts that translate into Google API actions (e.g., sending email, listing calendar events, reading Drive files, or searching YouTube). The server handles the OAuth flow and token management, while exposing a consistent MCP interface for the supported tools. The example configuration shows how to wire the server into an MCP-enabled client by specifying the npx-based invocation with the server URL.

Prerequisites:

• A Google Cloud project with OAuth credentials (Client ID and Client Secret) and enabled APIs for Gmail, Calendar, Drive, Tasks, YouTube, and People/Contacts
• A Cloudflare account and Wrangler available in your environment
• Node.js/npx (via Bun or npm) for local testing and deployment

Deployment steps:

1. Clone the repository and navigate to it

git clone https://github.com/vakharwalad23/google-mcp-remote.git
cd google-mcp-remote

2. Install dependencies

bun install

3. Configure secrets in Cloudflare (values sourced from Google Cloud Console)

wrangler secret put GOOGLE_OAUTH_CLIENT_ID
wrangler secret put GOOGLE_OAUTH_CLIENT_SECRET
wrangler secret put COOKIE_ENCRYPTION_KEY

4. Create a KV namespace for OAuth state (the ID will be used in wrangler.toml or wrangler.jsonc)

wrangler kv:namespace create OAUTH_KV

5. Deploy to Cloudflare Workers

bun run deploy

6. Note the deployment URL provided by Wrangler (e.g., https://your-project.your-username.workers.dev)

Local development and testing:

bun install
bun run dev

Test tooling locally with MCP Inspector or by pointing your MCP client to http://localhost:8788/sse if you run a local dev server.

• Ensure OAuth redirect URIs in Google Cloud Console include both the deployed Cloudflare Worker URL and the local testing URL (http://localhost:8788/callback).
• If you encounter OAuth issues, verify that the Google Cloud project has the required APIs enabled and that the OAuth client ID/secret match the deployed environment.
• For security, never share your Google OAuth credentials or the encryption key; rotate keys periodically and revoke access if a leak is suspected.
• The MCP URL to configure in AI clients should end with /sse to enable Server-Sent Events-based communication.
• When testing offline, you can use Cloudflare AI Playground with your deployed URL to verify functionality before integrating with an AI client.