Building MCP Server on Cloudflare

Build remote Model Context Protocol (MCP) servers on Cloudflare Workers with tools, OAuth authentication, and production deployment.

Download this Skill

npx degit LangbaseInc/agent-skills/building-mcp-server-on-cloudflare my-mcp-cloudflare

Overview

Deploy MCP servers to Cloudflare's global edge network for low-latency, scalable LLM tool integration.

Why Cloudflare Workers?

Global Edge - Deploy to 300+ locations worldwide
Zero Cold Starts - Instant execution
Auto-scaling - Handle any load
Low Cost - 100k requests/day free
KV/D1/R2 - Built-in storage options
OAuth - Easy authentication

Architecture

Claude/LLM → MCP Client → Cloudflare Worker (MCP Server)
                                    ↓
                          [KV, D1, R2, APIs]

Key Features

Tools

Expose functions to LLMs:

server.tool({
  name: 'get_weather',
  description: 'Get weather for a city',
  parameters: {
    city: { type: 'string' }
  },
  handler: async ({ city }) => {
    const weather = await fetchWeather(city);
    return { weather };
  }
});

Resources

Provide data access:

server.resource({
  uri: 'file://data.json',
  handler: async () => {
    return await env.KV.get('data');
  }
});

OAuth

Secure access:

const oauth = new OAuth({
  clientId: env.CLIENT_ID,
  clientSecret: env.CLIENT_SECRET,
});

Development Setup

1. Install Wrangler

npm install -g wrangler

2. Create Project

npm create cloudflare@latest my-mcp-server
cd my-mcp-server

3. Install MCP SDK

npm install @modelcontextprotocol/sdk

4. Develop Locally

wrangler dev

Worker Implementation

import { MCPServer } from '@modelcontextprotocol/sdk';

export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    const server = new MCPServer({
      name: 'my-mcp-server',
      version: '1.0.0',
    });

    // Register tools
    server.tool({
      name: 'example',
      description: 'Example tool',
      handler: async (args) => {
        // Tool logic here
        return { result: 'success' };
      },
    });

    return server.handleRequest(request);
  },
};

Storage Options

KV (Key-Value)

await env.MY_KV.put('key', 'value');
const value = await env.MY_KV.get('key');

D1 (SQL Database)

const result = await env.DB.prepare(
  'SELECT * FROM users WHERE id = ?'
).bind(userId).all();

R2 (Object Storage)

await env.MY_BUCKET.put('file.txt', data);
const file = await env.MY_BUCKET.get('file.txt');

Authentication

OAuth Flow

User authorizes app
Receive authorization code
Exchange for access token
Store in KV/D1
Use token for API calls

Deployment

1. Configure

# wrangler.toml
name = "my-mcp-server"
main = "src/index.ts"
compatibility_date = "2024-01-01"

[[kv_namespaces]]
binding = "MY_KV"
id = "your-kv-id"

2. Deploy

wrangler deploy

3. Set Secrets

wrangler secret put CLIENT_SECRET

Best Practices

Validate all inputs
Handle errors gracefully
Use TypeScript
Implement rate limiting
Cache responses
Monitor usage
Version your API
Document tools

Use Cases

API integration (Stripe, GitHub, etc.)
Database queries
File processing
Data transformation
External service calls
Webhook handling
Real-time data access

Cost Optimization

Use KV for caching
Batch database queries
Minimize external calls
Enable compression
Use Workers Analytics

Skills