Remote MCP Server Authless

You can run a remote MCP (Multi-Client Protocol) server that exposes its tools over HTTP for easy use with MCP clients, or run a local launcher that connects to a remote server. This setup lets you deploy, customize, and connect to your MCP server from playground clients or desktop tooling to execute your tools on demand.

How to use

Connect to your MCP server from an MCP client or the Cloudflare AI Playground to start using your tools. The remote server is reachable over HTTP at a URL like https://remote-mcp-server-authless.your-account.workers.dev/sse. You can also point a local proxy client at your running server to access the same tools from your desktop environment.

Using the Cloudflare AI Playground, you can interact with your MCP server without configuring a local client. Open the playground, enter your server URL, and begin issuing tool requests directly through the interface.

On your desktop, you can connect via a remote proxy by configuring your MCP client with the provided HTTP endpoint. For example, you can point a local launcher to the server’s SSE endpoint and use your tools as if they were running locally.

How to install

Prerequisites: ensure you have Node.js and npm installed on your machine. You can verify by running node -v and npm -v.

Step 1: Create or deploy the remote MCP server. You can deploy via a browser-based flow or run this command locally to bootstrap the server project.

Step 2: If you choose the browser-based deployment, open the deployment URL to initialize the MCP server project with the authless template. The deployment will create a server instance accessible at a URL like remote-mcp-server-authless.your-account.workers.dev/sse.

Step 3: Alternatively, run this command locally to create the project on your machine:

```bash
npm create cloudflare@latest -- my-mcp-server --template=cloudflare/ai/demos/remote-mcp-authless

This will scaffold a local MCP server ready to run.

## Additional content

Customizing your MCP server is straightforward: define each tool inside the init() method of `src/index.ts` using `this.server.tool(...)` to register your tools and make them available to MCP clients.

Connect from Cloudflare AI Playground to interact with your server without additional client setup. Use the server URL you deployed earlier to access the MCP tools directly from the playground.

If you prefer to use Claude Desktop or another local MCP client, configure your client to connect to the remote server. Update your client configuration to route tool calls to the server’s SSE endpoint, for example:\n

{
  "mcpServers": {
    "calculator": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "http://localhost:8787/sse"  // or remote-mcp-server-authless.your-account.workers.dev/sse
      ]
    }
  }
}

Restart Claude and you should see the tools become available.