Token导航 LogoToken导航TokenDH.com
前端设计敏感数据github未标认证来源可访问许可证需确认审计提醒

building-cloudflare-mcpbuilding Cloudflare MCP 命令行

Agent Skill

building-cloudflare-mcp 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

196

周安装

8

GitHub Stars

公开资料未说明

下载量

63
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:building-cloudflare-mcp(building Cloudflare MCP 命令行)
来源仓库:https://github.com/websmartteam/cor-code
仓库路径:skills/building-cloudflare-mcp
安装命令:
npx skills add https://github.com/websmartteam/cor-code --skill building-cloudflare-mcp
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/websmartteam/cor-code --skill building-cloudflare-mcp

简介

用于在 Cloudflare Workers 上部署 Model Context Protocol 服务器,实现低延迟全局接入。

  • 适用于需要边缘计算与 Claude API 集成的场景,支持无冷启动的持续服务。
  • 提供 MCP Connector 模式下的认证、路由与代码执行模板,兼容主流开发框架。
  • 安装使用 GitHub 仓库,建议检查项目维护状态并遵循官方文档进行本地测试验证。
  • building-cloudflare-mcp 属于前端设计类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Cloudflare MCP Connector

Build and deploy MCP (Model Context Protocol) servers on Cloudflare Workers using the MCP Connector pattern.

Why Cloudflare Workers for MCP?

  • Global edge deployment - Low latency worldwide
  • MCP Connector compatible - Works with Claude API MCP integration
  • No cold starts - Always-on serverless
  • Free tier - 100K requests/day free
  • Easy authentication - Headers-based auth

Official Documentation References

When you need deeper context, fetch these with WebFetch:

Code Mode (Advanced Pattern)

For high-scale deployments with many tools, consider the Code Mode pattern:

  • Present MCP tools as code APIs (filesystem structure)
  • Agent writes code to call tools instead of direct tool calls
  • 98.7% token savings for large tool sets
  • Filter/transform data in execution environment before returning

Why it works: LLMs have seen millions of real TypeScript examples in training, but only contrived synthetic tool-call examples.

Cloudflare Agents SDK (built-in Code Mode):

import { codemode } from "agents/codemode/ai";

const {system, tools} = codemode({
  system: "You are a helpful assistant",
  tools: { /* tool definitions */ },
});

const stream = streamText({
  model: openai("gpt-5"),
  system,
  tools,
  messages: [{ role: "user", content: "..." }]
});

Docs: https://github.com/cloudflare/agents/blob/main/docs/codemode.md

MCP Server Structure

Minimal Worker Template

// src/index.ts
export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    // CORS preflight
    if (request.method === 'OPTIONS') {
      return new Response(null, {
        headers: {
          'Access-Control-Allow-Origin': '*',
          'Access-Control-Allow-Methods': 'POST, OPTIONS',
          'Access-Control-Allow-Headers': 'Content-Type, Authorization',
        },
      });
    }

    // Auth check
    const authHeader = request.headers.get('Authorization');
    if (authHeader !== `Bearer ${env.MCP_API_KEY}`) {
      return Response.json({ error: 'Unauthorized' }, { status: 401 });
    }

    // Parse MCP request
    const body = await request.json() as MCPRequest;

    // Handle MCP methods
    switch (body.method) {
      case 'tools/list':
        return Response.json({
          jsonrpc: '2.0',
          id: body.id,
          result: { tools: getToolsList() }
        });

      case 'tools/call':
        const result = await handleToolCall(body.params, env);
        return Response.json({
          jsonrpc: '2.0',
          id: body.id,
          result
        });

      default:
        return Response.json({
          jsonrpc: '2.0',
          id: body.id,
          error: { code: -32601, message: 'Method not found' }
        });
    }
  },
};

interface MCPRequest {
  jsonrpc: '2.0';
  id: string | number;
  method: string;
  params?: Record<string, unknown>;
}

interface Env {
  MCP_API_KEY: string;
  // Add other bindings (KV, D1, R2, etc.)
}

function getToolsList() {
  return [
    {
      name: 'example_tool',
      description: 'Description of what this tool does',
      inputSchema: {
        type: 'object',
        properties: {
          param1: { type: 'string', description: 'First parameter' },
        },
        required: ['param1'],
      },
    },
  ];
}

async function handleToolCall(params: { name: string; arguments: Record<string, unknown> }, env: Env) {
  switch (params.name) {
    case 'example_tool':
      return { content: [{ type: 'text', text: `Result: ${params.arguments.param1}` }] };
    default:
      throw new Error(`Unknown tool: ${params.name}`);
  }
}

wrangler.toml

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

[vars]
# Non-secret config here

# Secrets added via: wrangler secret put MCP_API_KEY

Claude Code Configuration

Add Remote MCP Server

# Using CLI (user scope for global access)
claude mcp add --scope user my-mcp-server \
  --transport http \
  --url "https://my-mcp-server.username.workers.dev" \
  --header "Authorization: Bearer YOUR_API_KEY"

Manual Configuration (~/.claude.json)

{
  "mcpServers": {
    "my-mcp-server": {
      "type": "http",
      "url": "https://my-mcp-server.username.workers.dev",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Auto-Approve Tools (~/.claude/settings.json)

{
  "autoApproveTools": [
    "mcp__my-mcp-server__*"
  ]
}

Tool Naming Convention

Tools from MCP servers follow this pattern:

mcp__<server-name>__<tool-name>

Examples:

  • mcp__my-mcp-server__example_tool
  • mcp__supabase__execute_sql
  • mcp__context7__query-docs

Authentication Patterns

Bearer Token (Recommended)

const authHeader = request.headers.get('Authorization');
if (authHeader !== `Bearer ${env.MCP_API_KEY}`) {
  return Response.json({ error: 'Unauthorized' }, { status: 401 });
}

API Key Header

const apiKey = request.headers.get('X-API-Key');
if (apiKey !== env.API_KEY) {
  return Response.json({ error: 'Unauthorized' }, { status: 401 });
}

IP Allowlist (Additional Layer)

const clientIP = request.headers.get('CF-Connecting-IP');
const allowedIPs = env.ALLOWED_IPS?.split(',') || [];
if (allowedIPs.length && !allowedIPs.includes(clientIP)) {
  return Response.json({ error: 'Forbidden' }, { status: 403 });
}

Deployment Workflow

# 1. Create project
npm create cloudflare@latest my-mcp-server -- --template worker-typescript

# 2. Install dependencies
cd my-mcp-server
npm install

# 3. Add secrets
wrangler secret put MCP_API_KEY
# Enter your secure API key

# 4. Deploy
wrangler deploy

# 5. Add to Claude Code
claude mcp add --scope user my-mcp-server \
  --transport http \
  --url "https://my-mcp-server.username.workers.dev" \
  --header "Authorization: Bearer YOUR_API_KEY"

# 6. Restart Claude Code
# Exit and run: claude --resume

Advanced: Cloudflare Bindings

KV Storage

// wrangler.toml
[[kv_namespaces]]
binding = "MY_KV"
id = "abc123"

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

D1 Database

// wrangler.toml
[[d1_databases]]
binding = "DB"
database_name = "my-db"
database_id = "abc123"

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

R2 Storage

// wrangler.toml
[[r2_buckets]]
binding = "BUCKET"
bucket_name = "my-bucket"

// Usage
const object = await env.BUCKET.get('file.txt');
await env.BUCKET.put('file.txt', content);

Troubleshooting

MCP Not Loading

  1. Check worker is deployed: curl https://your-worker.workers.dev
  2. Verify auth header matches secret
  3. Restart Claude Code after config changes

Tools Not Appearing

  1. Check tools/list response format
  2. Verify tool schema is valid JSON Schema
  3. Check Claude Code logs: /mcp command

Permission Errors

  1. Add to autoApproveTools in settings.json
  2. Use wildcard: mcp__my-mcp-server__*

Security Checklist

  • Use strong, unique API keys (32+ chars)
  • Store secrets via wrangler secret put
  • Never commit secrets to git
  • Consider IP allowlisting for sensitive MCPs
  • Use HTTPS only (Cloudflare provides this)
  • Implement rate limiting if needed
  • Log access attempts for auditing

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

32.69%
按下载量换算21

Claude

30.37%
按下载量换算19

Cursor

18.39%
按下载量换算12

Gemini CLI

9.9%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills