Remote MCP Server On Cloudflare I68

This project demonstrates how to set up a remote MCP server on Cloudflare Workers, including OAuth login functionality, and supports connection and invocation of tool services through the MCP checker and Claude desktop client.

Artificial intelligence chatbots Developer tools #Remote Service #AI Tools #Protocol Development #Cloud Deployment Local .TypeScript

rating : 2 points

downloads : 10

update time : 2025-04-23

Core Features

Cloudflare AI Service provides high - performance AI models, supports real - time request processing and batch operations, and is suitable for various application scenarios.

Use Cases

This service is widely used in fields such as data analysis, automated processes, and intelligent customer service, helping developers quickly implement AI capabilities.

Advantages

High availability, low latency, strong scalability, support for multiple programming languages and integration methods, and provides comprehensive documentation and technical support.

Cloudflare DeploymentThe AI service is optimized based on Cloudflare's global network to ensure low latency and high availability.

Toolset SupportIt has built - in multiple AI tools, including functions such as mathematical calculations, text processing, and data analysis.

Security AssuranceIt uses end - to - end encryption and strict access control to ensure data security.

Initialize the Service

Create a Cloudflare account and enable the AI service.

Install Dependencies

Install necessary libraries and tools to support AI functions.

Configure API Key

Set the Cloudflare API key in the project for authentication.

Call the AI Service

Use the provided API interface to call AI functions.

Process the Response

Parse the result returned by the API and perform subsequent processing.

Mathematical CalculationUse the built - in mathematical tool for simple addition.

Text ProcessingPerform natural language understanding on the input text and generate a response.

Why do my requests always time out?

How to improve the processing speed?

What should I do if I encounter an error?

🚀 Running a Remote MCP Server on Cloudflare

Let's set up a remote MCP server and use OAuth for login!

🚀 Quick Start

💻 Local Development

# Clone the repository
git clone git@github.com:cloudflare/ai.git

# Install dependencies
cd ai
npm install

# Run locally
npx nx dev remote-mcp-server

You should be able to open a browser window at http://localhost:8787/.

🔗 Connecting the MCP Inspector to Your Server

To explore the new MCP API, you can use the MCP Inspector.

Start it using npx @modelcontextprotocol/inspector.
In the inspector, switch the transport type to SSE and enter http://localhost:8787/sse as the URL of the MCP server, then click "Connect".
You'll enter a (simulated) user/password login screen. Enter any email and password to log in.
You should be redirected back to the MCP Inspector and be able to list and call any defined tools!

🔗 Connecting Claude Desktop to Your Local MCP Server

The MCP Inspector is great, but we really want to connect it to Claude! Follow Anthropic's Quick Start Guide, go to Settings > Developer > Edit Configuration in Claude Desktop to find the configuration file. Open the file in a text editor and replace it with the following content:

{
  "mcpServers": {
    "math": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "http://localhost:8787/sse"
      ]
    }
  }
}

This will run a local proxy that allows Claude to communicate with your MCP server over HTTP. When you open Claude, a browser window should open and allow you to log in. You should see available tools in the bottom right corner. Depending on the correct prompts, Claude may request to call a tool.

🚀 Deployment to Cloudflare

npx wrangler kv namespace create OAUTH_KV
Follow the instructions to add the KV namespace ID to wrangler.jsonc.
npm run deploy

🔗 Calling Your Newly Deployed Remote MCP Server from a Remote MCP Client

Start it using npx @modelcontextprotocol/inspector.
In the inspector, switch the transport type to SSE and enter http://localhost:8787/sse as the URL of the MCP server, then click "Connect".
You'll enter a (simulated) user/password login screen. Enter any email and password to log in.
You should be redirected back to the MCP Inspector and be able to list and call any defined tools!

🔗 Configuring Claude Desktop

To connect Claude Desktop to your remote MCP server, edit its configuration file:

{
  "mcpServers": {
    "math": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "http://localhost:8787/sse"
      ]
    }
  }
}

This allows Claude to communicate with your remote MCP server over HTTP. When you open Claude, a browser window should open and allow you to log in. You should see available tools in the bottom right corner. Depending on the correct prompts, Claude may request to call a tool.