发现优秀的 MCP 服务器

cod-api MCP Server

Exposes REST API endpoints defined in an OpenAPI Specification as MCP tools, allowing AI models to call them via the ModelContext Protocol.

Weather MCP Demo

This MCP server provides tools to get current weather, forecasts, and weather alerts from the US National Weather Service via REST APIs, enabling AI agents to query live weather data.

MySQL MCP Server

Enables direct interaction with MySQL databases through SQL queries, table exploration, and schema inspection with support for prepared statements and connection pooling.

Minimal MCP Server

好的，以下是一个 Model Context Protocol (MCP) 服务器的最小实现，用 Python 编写，并使用 `asyncio` 库进行异步操作。 这个例子展示了如何监听连接，接收消息，并发送简单的响应。 ```python import asyncio import json async def handle_client(reader, writer): """处理单个客户端连接.""" addr = writer.get_extra_info('peername') print(f"连接来自 {addr}") try: while True: data = await reader.read(1024) # 读取最多 1024 字节 if not data: break message = data.decode() print(f"接收到消息: {message}") try: # 尝试解析 JSON request = json.loads(message) # 在这里添加你的 MCP 逻辑 # 例如，根据请求类型执行不同的操作 if "method" in request: method = request["method"] if method == "ping": response = {"result": "pong"} elif method == "get_model_info": response = {"result": {"model_name": "MyModel", "version": "1.0"}} # 示例模型信息 else: response = {"error": "Unknown method"} else: response = {"error": "Invalid request"} response_json = json.dumps(response) writer.write(response_json.encode()) await writer.drain() # 刷新缓冲区 print(f"发送响应: {response_json}") except json.JSONDecodeError: print("接收到无效的 JSON") writer.write(b'{"error": "Invalid JSON"}') await writer.drain() except ConnectionError as e: print(f"连接错误: {e}") finally: print(f"关闭连接 {addr}") writer.close() await writer.wait_closed() async def main(): """启动服务器.""" server = await asyncio.start_server( handle_client, '127.0.0.1', 8888) # 监听本地地址 127.0.0.1 端口 8888 addr = server.sockets[0].getsockname() print(f'服务于 {addr}') async with server: await server.serve_forever() if __name__ == "__main__": asyncio.run(main()) ``` **代码解释:** 1. **`handle_client(reader, writer)` 函数:** - 这是处理每个客户端连接的主要函数。 - `reader` 和 `writer` 是 `asyncio.StreamReader` 和 `asyncio.StreamWriter` 对象，用于读取和写入数据。 - `writer.get_extra_info('peername')` 获取客户端的地址。 - `reader.read(1024)` 从客户端读取最多 1024 字节的数据。 - `data.decode()` 将接收到的字节数据解码为字符串。 - `json.loads(message)` 尝试将接收到的消息解析为 JSON 对象。 - **MCP 逻辑:** 这是代码的核心部分，你需要根据 MCP 协议的要求实现你的逻辑。 在这个例子中，它检查请求中是否存在 "method" 字段，并根据方法名称返回不同的响应。 `ping` 方法返回 `pong`，`get_model_info` 返回一个示例模型信息。 - `json.dumps(response)` 将响应转换为 JSON 字符串。 - `writer.write(response_json.encode())` 将 JSON 字符串编码为字节数据并发送给客户端。 - `await writer.drain()` 刷新缓冲区，确保数据被发送。 - `writer.close()` 关闭连接。 - `await writer.wait_closed()` 等待连接完全关闭。 - 异常处理：包括 `json.JSONDecodeError` 处理无效的 JSON，以及 `ConnectionError` 处理连接错误。 2. **`main()` 函数:** - `asyncio.start_server(handle_client, '127.0.0.1', 8888)` 启动一个 TCP 服务器，监听本地地址 127.0.0.1 的 8888 端口。 `handle_client` 函数将处理每个新的客户端连接。 - `server.serve_forever()` 运行服务器，直到手动停止。 3. **`if __name__ == "__main__":` 块:** - 确保 `asyncio.run(main())` 只在脚本直接运行时执行，而不是作为模块导入时执行。 **如何运行:** 1. **保存代码:** 将代码保存为 `mcp_server.py`。 2. **运行脚本:** 在终端中运行 `python mcp_server.py`。 **如何测试:** 你可以使用 `telnet` 或 `netcat` 等工具来连接服务器并发送 MCP 请求。 例如： ```bash telnet 127.0.0.1 8888 ``` 然后，你可以发送一个 JSON 请求，例如： ```json {"method": "ping"} ``` 服务器应该返回： ```json {"result": "pong"} ``` 或者发送： ```json {"method": "get_model_info"} ``` 服务器应该返回： ```json {"result": {"model_name": "MyModel", "version": "1.0"}} ``` **重要注意事项:** * **错误处理:** 这个例子只包含基本的错误处理。 在实际应用中，你需要添加更完善的错误处理机制，例如记录错误日志，返回更详细的错误信息给客户端。 * **安全性:** 这个例子没有考虑安全性。 在生产环境中，你需要采取安全措施，例如身份验证、授权、加密等。 * **MCP 协议:** 这个例子只是一个框架。 你需要根据 MCP 协议的规范来实现你的 MCP 逻辑。 这包括定义消息格式、方法名称、参数、返回值等。 * **异步编程:** `asyncio` 是一个强大的异步编程库。 如果你不熟悉 `asyncio`，建议先学习一下 `async` 和 `await` 的用法。 * **依赖:** 这个例子只依赖于 Python 的标准库 `asyncio` 和 `json`，不需要安装额外的依赖。 * **扩展性:** 这个例子是一个最小实现。 你可以根据需要扩展它，例如添加多线程支持、使用更高效的序列化方法、支持更多的 MCP 方法等。 **下一步:** 1. **定义你的 MCP 协议:** 明确你的 MCP 协议的规范，包括消息格式、方法名称、参数、返回值等。 2. **实现你的 MCP 逻辑:** 根据你的 MCP 协议，实现 `handle_client` 函数中的 MCP 逻辑。 3. **添加错误处理:** 添加更完善的错误处理机制。 4. **考虑安全性:** 采取安全措施，例如身份验证、授权、加密等。 5. **测试和调试:** 充分测试和调试你的服务器。 这个最小实现为你提供了一个起点。 你可以根据你的具体需求来扩展和修改它。

Shopify MCP Server

Enables interaction with Shopify store data through the GraphQL API, providing tools for managing products, customers, orders, and collections.

OmniFocus MCP Server

An MCP server connecting OmniFocus to AI assistants, enabling task and project management via natural language.

Agentic HEC-RAS

MCP server that enables natural-language-driven, headless HEC-RAS modeling, letting agents inspect, edit, run, compare, and plot 1D/2D hydraulic models without the GUI.

autor-texto-seguro

Enables secure text authoring and privacy-focused document processing through MCP. Provides tools for style analysis, conservative text revision, comparison with author samples, and removal of metadata from DOCX/PDF files.

crosscheck

MCP server that automatically cross-checks claims from Reddit posts against trusted news sources and delivers a daily digest to Discord, without making final truth judgments.

raindrop-mcp

Raindrop.io (书签服务) 的 MCP 服务器

getstoreready-mcp

Enables AI clients to manage app store projects via the GetStoreReady API, including creating projects, uploading screenshots, applying templates, updating listings, and checking project status.

Claude Code DingTalk MCP Server

Integrates Claude Code with DingTalk (钉钉) robot notifications, allowing users to send task completion alerts and various message formats to DingTalk groups from Claude Code.

screencye

Eyes for text-only LLMs: decodes screenshots into exact structured text (words, coordinates, sizes, colors) using pure-code CV and OCR. Enables text-only models to reason about UI layouts without vision models or VRAM usage.

魂斗罗 MCP Demo

Enables creating a Contra-style game session that can be played inline in MCP-compatible clients or via a browser fallback URL.

PrestaShop MCP Server

Provides instant, offline access to comprehensive PrestaShop development documentation including 647+ hooks, module guides, component architecture, APIs, and theme development resources for AI assistants.

chatlab-mcp

MCP server for ChatLab that automates the chatlab-cli in headless HTTP mode, enabling session management, message retrieval, SQL queries, and API calls via natural language.

VMware Workstation Pro MCP Server

Enables controlling VMware Workstation Pro virtual machines, including listing, power management, snapshots, cloning, guest operations, networking, and screenshots.

LiteAPI MCP Server

Enables interaction with LiteAPI's hotel booking and travel services through MCP tools, supporting search, booking, vouchers, analytics, and more.

HAI-MCP

Provides a human-agent interface control plane for managing agent state, focus, and next steps across any MCP-compatible client, without calling an LLM itself.

MCP JSON Database Server

A JSON-based database MCP server with JWT authentication that enables user management, project tracking, department analysis, meeting management, and equipment tracking. Integrates with Claude Desktop to provide secure CRUD operations and analytics through natural language commands.

MCP Weather Project

An MCP server that provides real-time weather alerts for US states using the National Weather Service (NWS) API. It also includes utility features for message echoing and generating customizable greeting prompts.

veyra-forms

Form builder and response collector for AI agents. Reads are free, writes require Veyra commit mode.

escape-room-hub

This MCP server enables querying normalized escape room information and reservation availability via stdio or Streamable HTTP transports. It is a policy-driven scaffold that currently uses synthetic data only and does not perform real web crawling.

Web Search Tool

Provides a Google search tool via the Serper API, returning structured results and a summary answer for LLMs like Claude.

WebdriverIO MCP Server

A Model Context Protocol server that acts as a bridge, allowing LLMs and AI agents to control a web browser using WebdriverIO.

helios-calcom-mcp

Secure MCP wrapper for Cal.com integration, exposing calendar booking tools to Hermes/Helios.

context-bridge

Provides LLMs with secure, read-only access to local documentation by scanning directories, extracting content from PDF, DOCX, Markdown, and text files, and performing keyword searches.

leadforge

Enables lightweight CRM pipeline management with email sequences, project scanning, and MCP-based interaction for AI agents.

mcp-clickhouse

An MCP server for ClickHouse with enhanced filtering for database and table discovery, supporting LIKE/NOT LIKE patterns and both ClickHouse and chDB tools.