Rapid Deployment of MCP Services in Cloudflare Workers

General Introduction

Cloudflare Workers MCP is an open source project developed by Cloudflare and hosted on GitHub, designed to help developers quickly build and deploy Cloudflare Workers-based MCP (Model Context Protocol) servers. This tool combines the powerful cloud computing capabilities of Cloudflare Workers with the MCP protocol , allowing users to create custom tools through simple TypeScript functions , and integrated into MCP-enabled clients (such as the Anthropic (used form a nominal expression) Claude Desktop). It is especially suitable for developers who want to utilize Cloudflare infrastructure to realize efficient and lightweight server functions, supporting services such as R2, D1, KV, etc. The deployment process is simple, and the development experience is smooth, which is welcomed by the technical community.

Function List

• MCP Server Quick Setup: Rapidly spawn and deploy MCP servers to Cloudflare Workers via the CLI tool.
• Support for TypeScript development: Write function functions using TypeScript and generate tool descriptions using JSDoc annotations.
• Integrating Cloudflare Services: Supports R2 (object store), D1 (database), KV (key-value store) and other function calls.
• Seamless connectivity with MCP clients: Interfaces with MCP clients such as Claude Desktop for customization.
• Automatic Document Generation: Automatically generate LLM-friendly tool documentation from JSDoc comments in code.
• Efficient deployment process: One-click deployment to Cloudflare's global network in conjunction with the Wrangler CLI.
• Local Development Support: Provide a local test environment for debugging and iteration.

Using Help

Installation process

To use Cloudflare Workers MCP, you need to prepare some basic environment, then follow the steps below to install and configure the project. The whole process is straightforward and suitable for users with basic development experience.

1. Environmental preparation

• Node.js: Make sure you have Node.js installed on your computer (version >= 16.17.0), which can be done with the command node -v Check the version.
• Wrangler CLI: This is the official command line tool for Cloudflare Workers to deploy and manage Worker projects. Installation method:

  npm install -g wrangler

After installation, run wrangler login Log in to your Cloudflare account.

• Claude Desktop (optional): If you want to integrate the MCP server with Claude Desktop, you need to download and install the Claude Desktop application (official website: https://claude.ai/download).
• Git: to clone a project from GitHub, run the git --version Check if it is installed.

2. Cloning projects

Open a terminal and enter the following command to clone the Cloudflare Workers MCP repository:

git clone https://github.com/cloudflare/workers-mcp.git
cd workers-mcp

3. Installation of dependencies

Once in the project directory, install the necessary dependency packages. You can choose either npm or pnpm:

npm install

maybe

pnpm install

Once the installation is complete, the project will contain all the necessary libraries, such as the @modelcontextprotocol/sdk cap (a poem) tsxThe

4. Configuring Wrangler

Make sure your wrangler.toml The file contains Cloudflare account information. Run the following command to log in and generate the configuration file:

wrangler login

After successful login, edit wrangler.tomlAdd your account_id(available in the Cloudflare dashboard):

account_id = "Your account ID"

5. Local operations

Test the MCP server locally, running:

pnpm build
npx workers-mcp secret generate

This generates a key and stores it in the .dev.vars file. Then start the local development environment:

wrangler dev

The local URL of the access prompt (usually http://localhost:8787), check that the server is functioning properly.

6. Deployment to Cloudflare

After testing for errors, deploy to Cloudflare Workers using the following command:

pnpm deploy:worker

After a successful deployment, you'll get a URL for the Cloudflare Worker (e.g. https://your-worker.workers.dev), which is your MCP server address.

Functional operation flow

Key Features: Rapidly build MCP servers

1. Writing TypeScript Functions
   show (a ticket) src/index.tsin MyWorker class to add custom functions. Example:

   /**
   * Returns a friendly greeting.
   * @param name {string} The user's name.
   * @return {string} The content of the greeting.
   */
   sayHello(name: string) {
   return `Hello, ${name}! Greetings from MCP Worker! `;
   }
   

   JSDoc annotations automatically generate tool descriptions.

2. Generate Documentation
   (of a computer) run pnpm buildThe tool generates the dist/docs.json, contains description and parameter information for all tools.
3. test function
   run locally wrangler devIf you want to test it, you can do so by using a browser or curl test:

   curl http://localhost:8787
   

   Alternatively, configure the MCP server address directly in Claude Desktop by calling the sayHello Tools.

Feature: Integration with Claude Desktop

1. Configuring Claude Desktop
   Open the Claude Desktop configuration file claude_desktop_config.json(usually located in the user directory), add the MCP server:

   {
   "mcpServers": {
   "my-worker": {
   "command": "npx".
   "args": ["workers-mcp", "proxy", "https://your-worker.workers.dev"]
   }
   }
   }
   

   Save and restart Claude Desktop.

2. invocation tool
   In Claude Desktop, type something like "Please invoke the sayHello utility to greet John" and you will see the return result: "Hello John! Greetings from MCP Worker!".

Feature: Integration with Cloudflare Services

1. Using KV Storage
   exist wrangler.toml Add a KV binding to the

   [[kv_namespaces]]
   binding = "MY_KV"
   id = "Your KV ID"
   

   Manipulate KV in code:

   async setValue(key: string, value: string) {
   await this.env.MY_KV.put(key, value);
   return `Stored ${key}: ${value}`; }
   }
   

2. Deployment and testing
   After redeploying, call the setValue tool to verify that the data is deposited into the KV.

caveat

• Key Management: Run npx workers-mcp secret upload Upload your keys to Cloudflare to ensure security.
• Debugging Tips: Use wrangler dev --remote Connectable to production environment bindings for easy debugging.
• Extended functionality: Additional Cloudflare services (e.g., R2, D1) can be introduced by simply adding the following to the wrangler.toml Configure the binding in the

With these steps, you can quickly get started with Cloudflare Workers MCP and build a powerful MCP server!