cf-worker -template

Một skeleton đầy đủ để bạn bắt đầu nhanh với Cloudflare Workers: routes, utils, services, bindings (KV/D1/R2), Durable Objects, cron.

Installation

Run this command in your terminal to add the MCP server to Claude Code.

Run in terminal:

View docs

Command

claude mcp add --transport stdio pnstack-cf-worker-mcp-template node src/index.ts \
  --env CF_API_TOKEN="Cloudflare API token if deploying via Wrangler" \
  --env CF_ACCOUNT_ID="Cloudflare account ID" \
  --env WRANGLER_CONFIG="Optional: path to wrangler.toml or Wrangler config"

How to use

This MCP server template provides a Cloudflare Workers-based implementation that exposes an MCP (Streamable HTTP) endpoint at POST /mcp. It includes tools scaffolding (e.g., hello and api tools) that can be scaffolded and invoked through the MCP protocol. The template also demonstrates how to enable OpenAI-compatible embeddings and various Cloudflare bindings (KV, D1, R2, Durable Objects, and AI integrations). To interact with the MCP, initialize the session with a JSON-RPC 2.0 payload, list available tools, and call a specific tool by name with its arguments. If you request an event-stream response (Accept: text/event-stream), the server will emit SSE data for real-time updates during tool execution.

How to install

Prerequisites:

Node.js v18 or later
npm (comes with Node.js)
Wrangler CLI for Cloudflare Workers (optional if deploying directly with other tooling): npm i -g wrangler

Installation steps:

Clone or download the repository template.
Install dependencies: npm install
(Optional) Configure Wrangler for Cloudflare Workers by creating or editing wrangler.toml with your account and zone details.
Run in development mode: npm run dev
Deploy to Cloudflare Workers: npm run deploy

Notes:

If you plan to use bindings (KV, D1, R2, DO, AI), ensure they are configured in wrangler.toml and enabled in the Cloudflare dashboard.
Set necessary secrets such as CF_API_TOKEN and CF_ACCOUNT_ID when deploying or running authenticated operations.

Additional notes

Tips and common considerations:

The template supports multiple MCP tools (e.g., hello, api). Use the tool scaffold and tool invocation patterns shown in the README to add or call tools via JSON-RPC.
When enabling OpenAI embeddings, provide OPENAI_API_KEY as an environment variable and use the /v1/embeddings endpoints through the server's route (with proper baseURL pointing to the local or remote endpoint).
If you do not enable certain bindings (KV/D1/R2/DO), keep the related sections commented in wrangler.toml to avoid deployment errors.
For local testing, you can simulate the MCP calls using standard JSON-RPC payloads shown in the README (initialize, tools/list, tools/call).
Ensure compatibility_date is updated periodically to use the latest API surface.
If using SSE responses, set Accept: text/event-stream on the request and handle streaming data on the client side.

Related MCP Servers

mcp

114

MCP server for the Cloudflare API

MiniMax -JS

105

Official MiniMax Model Context Protocol (MCP) JavaScript implementation that provides seamless integration with MiniMax's powerful AI capabilities including image generation, video generation, text-to-speech, and voice cloning APIs.

MCP -Starter

A Model Context Protocol server starter template

ez

The easiest path to getting an MCP server going

google -remote

Collection of Google-native tools (e.g., Gmail, Calendar) for the MCP

canlii

canlii mcp for people in law in canada