Remote MCP Server on Cloudflare (Without Auth)

This example enables you to deploy a non-authenticated remote Model Context Protocol (MCP) server on Cloudflare Workers, facilitating easy integration with various AI applications.

Overview: What is a Remote MCP Server?

The Remote MCP Server on Cloudflare allows developers and AI model integrators to create a publicly accessible endpoint for their models or tools. This server adheres to the MCP protocol, ensuring seamless interaction with clients like Claude Desktop, Continue, Cursor, and more, without requiring authentication.

🔧 Core Features & MCP Capabilities

The core focus of this remote MCP server lies in its ability to provide an unauthenticated service on Cloudflare Workers. This setup is ideal for scenarios where rapid deployment and broad accessibility are necessary. Key capabilities include:

• Ease of Deployment: Users can deploy the server with a click using the provided button or through CLI commands.
• Tool Integration: Customize your MCP server by defining tools within init() method in src/index.ts.
• Client Connectivity: Connect directly from Cloudflare’s AI Playground, or utilize the mcp-remote proxy to link local clients like Claude Desktop.

⚙️ MCP Architecture & Protocol Implementation

The architecture leverages Cloudflare Workers for hosting, ensuring minimal latency and high performance. The implementation of the protocol focuses on:

1. Server Initialization: Tools are added via the this.server.tool(...) method within init().
2. Client Requests Handling: The server processes requests according to the MCP protocol.

MCP uses a publish-subscribe model for handling asynchronous communication (using Server-Sent Events, SSE). This diagram illustrates the flow:

graph TD
    A[AI Application] -->|MCP Client| B[MCP Protocol]
    B --> C[MCP Server]
    C --> D[Data Source/Tool]
    style A fill:#e1f5fe
    style C fill:#f3e5f5
    style D fill:#e8f5e8

🚀 Getting Started with Installation

Deployment through Cloudflare Workers:

Users can quickly launch the server by clicking this button:

After deployment, your URL will resemble: remote-mcp-server-authless.<your-account>.workers.dev/sse.

Local Machine Deployment:

Alternatively, install the server on a local machine with:

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

💡 Key Use Cases in AI Workflows

1. Data Aggregation for Analytical Tools: Deploy tools that gather and process data remotely, ensuring secure and scalable data handling.
2. Integration with Custom Applications: Quickly add custom-built models or tools to existing workflows without complex authentication mechanisms.

🔌 Integration with MCP Clients

Connection to Cloudflare AI Playground

1. Access the Cloudflare AI Playground.
2. Enter the generated URL (remote-mcp-server-authless.<your-account>.workers.dev/sse) in the connection dialog.
3. Your tools will be readily available for use within the playground.

Connecting with Claude Desktop

Follow the steps provided by Anthropic:

1. Navigate to Anthropic's Quickstart.
2. In Claude Desktop, go to Settings > Developer > Edit Config.
3. Add your remote MCP server setup:

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

📊 Performance & Compatibility Matrix

MCP Client	Resources	Tools	Prompts	Status
Claude Desktop	✅	✅	✅	Full Support
Continue	✅	✅	✅	Full Support
Cursor	❌	✅	❌	Tools Only

🛠️ Advanced Configuration & Security

While the server is designed for ease of use, some advanced configurations are available:

• API Key Environment Variable: Securely manage access by setting environment variables such as API_KEY.
• Custom Headers: Customize client-server communication through headers.

Example configuration:

{
  "mcpServers": {
    "[server-name]": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-[name]"],
      "env": {
        "API_KEY": "your-api-key"
      }
    }
  }
}

❓ Frequently Asked Questions (FAQ)

1. Q: Can I deploy this server on other platforms?

   • A: Yes, while Cloudflare Workers are used for quick deployment and ease of use, the server can be hosted elsewhere by replicating its logic.

2. Q: Does this server support authentication?

   • A: No, the provided example is specifically designed for unauthenticated access, but authentication mechanisms can be added via custom code.

3. Q: How does this impact privacy and security?

   • A: Privacy and security are maintained through careful configuration of environment variables and client communication.

4. Q: Can I integrate existing applications with this server?

   • A: Absolutely, use the provided configurations to easily add your tools or data sources.

5. Q: How can I manage multiple servers efficiently?

   • A: Leverage environment variables for different settings per server and consider using a centralized management system like GitHub Actions or Terraform.

👨‍💻 Development & Contribution Guidelines

Contributors are encouraged to follow these guidelines:

1. Branching: Develop new features on separate branches.
2. Testing: Write unit tests before submitting PRs.
3. Documentation: Ensure all changes have corresponding documentation updates.

🌐 MCP Ecosystem & Resources

Explore more resources and tools related to the Model Context Protocol:

• MCP Quickstart Guide
• MCP Client Libraries
• MCP Community Forum

By leveraging this remote MCP server, developers can significantly enhance the capabilities of their AI applications through seamless integration with various tools and data sources.