openai/building-mcp-server-on-cloudflare

Building MCP Servers on Cloudflare

Your knowledge of the MCP SDK and Cloudflare Workers integration may be outdated. Prefer retrieval over pre-training for any MCP server task.

Retrieval Sources

| Source | How to retrieve | Use for | |--------|----------------|---------| | MCP docs | https://developers.cloudflare.com/agents/mcp/ | Server setup, auth, deployment | | MCP spec | https://modelcontextprotocol.io/ | Protocol spec, tool/resource definitions | | Workers docs | Search tool or https://developers.cloudflare.com/workers/ | Runtime APIs, bindings, config |

When to Use

• User wants to build a remote MCP server
• User needs to expose tools via MCP
• User asks about MCP authentication or OAuth
• User wants to deploy MCP to Cloudflare Workers

Prerequisites

• Cloudflare account with Workers enabled
• Node.js 18+ and npm/pnpm/yarn
• Wrangler CLI (npm install -g wrangler)

Quick Start

Option 1: Public Server (No Auth)

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

Server runs at http://localhost:8788/mcp

Option 2: Authenticated Server (OAuth)

npm create cloudflare@latest -- my-mcp-server \
  --template=cloudflare/ai/demos/remote-mcp-github-oauth
cd my-mcp-server

Requires OAuth app setup. See references/oauth-setup.md.

Core Workflow

Step 1: Define Tools

Tools are functions MCP clients can call. Define them using server.tool():

import { McpAgent } from "agents/mcp";
import { z } from "zod";

export class MyMCP extends McpAgent {
  server = new Server({ name: "my-mcp", version: "1.0.0" });

  async init() {
    // Simple tool with parameters
    this.server.tool(
      "add",
      { a: z.number(), b: z.number() },
      async ({ a, b }) => ({
        content: [{ type: "text", text: String(a + b) }],
      })
    );

    // Tool that calls external API
    this.server.tool(
      "get_weather",
      { city: z.string() },
      async ({ city }) => {
        const response = await fetch(`https://api.weather.com/${city}`);
        const data = await response.json();
        return {
          content: [{ type: "text", text: JSON.stringify(data) }],
        };
      }
    );
  }
}

Step 2: Configure Entry Point

Public server (src/index.ts):

import { MyMCP } from "./mcp";

export default {
  fetch(request: Request, env: Env, ctx: ExecutionContext) {
    const url = new URL(request.url);
    if (url.pathname === "/mcp") {
      return MyMCP.serveSSE("/mcp").fetch(request, env, ctx);
    }
    return new Response("MCP Server", { status: 200 });
  },
};

export { MyMCP };

Authenticated server — See references/oauth-setup.md.

Step 3: Test Locally

# Start server
npm start

# In another terminal, test with MCP Inspector
npx @modelcontextprotocol/inspector@latest
# Open http://localhost:5173, enter http://localhost:8788/mcp

Step 4: Deploy

npx wrangler deploy

Server accessible at https://[worker-name].[account].workers.dev/mcp

Step 5: Connect Clients

Codex MCP client setup:

codex mcp add my-server -- npx mcp-remote https://my-mcp.workers.dev/mcp

Restart Codex after updating the MCP configuration.

Tool Patterns

Return Types

// Text response
return { content: [{ type: "text", text: "result" }] };

// Multiple content items
return {
  content: [
    { type: "text", text: "Here's the data:" },
    { type: "text", text: JSON.stringify(data, null, 2) },
  ],
};

Input Validation with Zod

this.server.tool(
  "create_user",
  {
    email: z.string().email(),
    name: z.string().min(1).max(100),
    role: z.enum(["admin", "user", "guest"]),
    age: z.number().int().min(0).optional(),
  },
  async (params) => {
    // params are fully typed and validated
  }
);

Accessing Environment/Bindings

export class MyMCP extends McpAgent<Env> {
  async init() {
    this.server.tool("query_db", { sql: z.string() }, async ({ sql }) => {
      // Access D1 binding
      const result = await this.env.DB.prepare(sql).all();
      return { content: [{ type: "text", text: JSON.stringify(result) }] };
    });
  }
}

Authentication

For OAuth-protected servers, see references/oauth-setup.md.

Supported providers:

• GitHub
• Google
• Auth0
• Stytch
• WorkOS
• Any OAuth 2.0 compliant provider

Wrangler Configuration

Minimal wrangler.toml:

name = "my-mcp-server"
main = "src/index.ts"
compatibility_date = "2024-12-01"

[durable_objects]
bindings = [{ name = "MCP", class_name = "MyMCP" }]

[[migrations]]
tag = "v1"
new_classes = ["MyMCP"]

With bindings (D1, KV, etc.):

[[d1_databases]]
binding = "DB"
database_name = "my-db"
database_id = "xxx"

[[kv_namespaces]]
binding = "KV"
id = "xxx"

Common Issues

"Tool not found" in Client

1. Verify tool name matches exactly (case-sensitive)
2. Ensure init() registers tools before connections
3. Check server logs: wrangler tail

Connection Fails

1. Confirm endpoint path is /mcp
2. Check CORS if browser-based client
3. Verify Worker is deployed: wrangler deployments list

OAuth Redirect Errors

1. Callback URL must match OAuth app config exactly
2. Check GITHUB_CLIENT_ID and GITHUB_CLIENT_SECRET are set
3. For local dev, use http://localhost:8788/callback

References

• references/examples.md — Official templates and production examples
• references/oauth-setup.md — OAuth provider configuration
• references/troubleshooting.md — Error codes and fixes