Remote Serverless MCP Servers

You want to run a remote MCP server, but you don’t want to host or pay for a server. Well you’re in luck, you can use the Vercel Hobby Tier to run it for free.

This is an adapter library that provides a stateless Python MCP server that runs on Vercel Functions, enabling serverless deployment of MCP tools as edge functions with self-hosting installers.

Prerequisites

Fork this repo and connect it to your Vercel account. Once it’s deployed get the URL (your-domain.vercel.app).

If you don’t have uv installed, install it from astral:

curl -LsSf https://astral.sh/uv/install.sh | sh

Then, install the MCP server using the self-hosted installer:

uv run https://your-domain.vercel.app/install.py

And we’re done. This automatically:

• Configures Claude Desktop and Cursor to use the MCP server
• Uses the bridge script served directly from the server
• No downloads or local files needed

Go to your client and type something stupid like “Use MCP to get the server time” and it should use the remote server.

Authentication (Optional)

The server supports optional API key authentication to restrict access. To enable authentication:

1. Server-side: Set the MCP_API_KEY environment variable in your Vercel deployment
2. Client-side: The installer will prompt for an API key during setup

You can set the API key in your vercel.json file or in the Vercel dashboard:

{
  "env": {
    "MCP_API_KEY": "your-secret-key-here"
  }
}

When an API key is configured on the server, all MCP requests must include the key in the X-API-Key header. The bridge script automatically handles this when the MCP_API_KEY environment variable is set in the client configuration.

Local Development

uv sync
vercel dev

Deploy

You just push to Github and it should just work.

Or manually deploy:

vercel --prod