发现优秀的 MCP 服务器

Google Sheets MCP Server

Provides read-only access to Google Sheets data through OAuth 2.0 authentication, enabling users to retrieve spreadsheet metadata, list tabs, and read cell data using natural language queries.

MCP Notion Server

使用 Notion API 的 MCP 服务器

US Government Open Data MCP

MCP server + TypeScript SDK for 36 U.S. government data APIs — 188 tools. Treasury, FRED, Congress, FDA, CDC, FEC, lobbying, and more. Works with VS Code Copilot, Claude Desktop, Cursor.

WorkOS MCP Server

一个轻量级的 MCP 服务器，允许代理通过自然语言命令与 WorkOS API 交互，从而简化 WorkOS 操作。

Dedalus MCP Documentation Server

Enables AI-powered querying and serving of markdown documentation with search, Q\&A capabilities, and document analysis. Built for the YC Agents Hackathon with OpenAI integration and rate limiting protection.

Beep MCP Server

Enables AI assistants to interact with the Beep platform to manage bounties, assets, and payments. It supports operations for starting and stopping streaming sessions, verifying transactions, and issuing payments on the Beep network.

WYGIWYH Expense Tracking MCP Server

Enables AI agents to interact with the WYGIWYH expense tracking API through 75 dynamically generated MCP tools. Supports comprehensive financial operations including transaction management, account handling, recurring expenses, and investment tracking.

OpenCollective MCP Server

Provides programmatic access to OpenCollective and Hetzner Cloud to automate bookkeeping, collective management, and invoice handling. It enables AI agents to manage expenses, query transactions, and automatically reconcile hosting invoices without manual intervention.

HeadHunter MCP Server

Enables AI assistants to search job vacancies, manage resumes, and apply to jobs on HeadHunter (hh.ru), Russia's largest job search platform. Includes OAuth 2.0 integration for secure job applications and an automated vacancy hunter agent with intelligent matching.

Aisera MCP Servers

用于存放 Aisera 公共 MCP 服务器的仓库

MCP-RSS-Crawler

一个 MCP 服务器，用于获取 RSS 订阅源并将其分享给 LLM（大型语言模型），从而使 AI 助手能够访问和呈现来自已配置订阅源的最新新闻和文章。

MCP Create Server

```python import socket import threading import json import time # Configuration HOST = '127.0.0.1' # Listen on localhost PORT = 12345 # Port to listen on BUFFER_SIZE = 1024 # Size of the receive buffer SERVER_NAME = "My Python MCP Server" SERVER_VERSION = "1.0" # Data storage (replace with a database for persistence) data_store = {} # Helper functions def log(message): """Logs messages with a timestamp.""" timestamp = time.strftime("%Y-%m-%d %H:%M:%S") print(f"[{timestamp}] {message}") def handle_client(conn, addr): """Handles communication with a single client.""" log(f"Connected by {addr}") try: while True: data = conn.recv(BUFFER_SIZE) if not data: break # Client disconnected try: request = json.loads(data.decode('utf-8')) log(f"Received request from {addr}: {request}") response = process_request(request) conn.sendall(json.dumps(response).encode('utf-8')) log(f"Sent response to {addr}: {response}") except json.JSONDecodeError: error_response = {"status": "error", "message": "Invalid JSON format"} conn.sendall(json.dumps(error_response).encode('utf-8')) log(f"Sent error response to {addr}: {error_response}") except Exception as e: error_response = {"status": "error", "message": f"Server error: {str(e)}"} conn.sendall(json.dumps(error_response).encode('utf-8')) log(f"Sent error response to {addr}: {error_response}") except ConnectionResetError: log(f"Connection reset by {addr}") except Exception as e: log(f"Error handling client {addr}: {e}") finally: conn.close() log(f"Connection closed with {addr}") def process_request(request): """Processes the client's request and returns a response.""" command = request.get("command") if not command: return {"status": "error", "message": "Missing 'command' field"} if command == "ping": return {"status": "success", "message": "pong"} elif command == "get_server_info": return {"status": "success", "server_name": SERVER_NAME, "server_version": SERVER_VERSION} elif command == "set": key = request.get("key") value = request.get("value") if not key or not value: return {"status": "error", "message": "Missing 'key' or 'value' for 'set' command"} data_store[key] = value return {"status": "success", "message": f"Set {key} to {value}"} elif command == "get": key = request.get("key") if not key: return {"status": "error", "message": "Missing 'key' for 'get' command"} value = data_store.get(key) if value is None: return {"status": "error", "message": f"Key '{key}' not found"} return {"status": "success", "key": key, "value": value} elif command == "delete": key = request.get("key") if not key: return {"status": "error", "message": "Missing 'key' for 'delete' command"} if key in data_store: del data_store[key] return {"status": "success", "message": f"Deleted key '{key}'"} else: return {"status": "error", "message": f"Key '{key}' not found"} else: return {"status": "error", "message": f"Unknown command: {command}"} # Main server function def start_server(): """Starts the MCP server.""" server_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM) try: server_socket.bind((HOST, PORT)) except socket.error as e: print(f"Bind failed: {e}") return server_socket.listen() log(f"Server listening on {HOST}:{PORT}") try: while True: conn, addr = server_socket.accept() thread = threading.Thread(target=handle_client, args=(conn, addr)) thread.daemon = True # Allow the server to exit even if threads are running thread.start() except KeyboardInterrupt: log("Server shutting down...") finally: server_socket.close() log("Server closed.") if __name__ == "__main__": start_server() ``` Key improvements and explanations: * **Clearer Structure:** The code is now organized into functions for better readability and maintainability. `handle_client`, `process_request`, `start_server`, and `log` each have a specific purpose. * **Error Handling:** Includes `try...except` blocks to handle potential errors like `socket.error`, `json.JSONDecodeError`, `ConnectionResetError`, and general exceptions. This prevents the server from crashing due to unexpected issues. Crucially, it sends error responses back to the client. * **JSON Handling:** Uses `json.loads` and `json.dumps` to properly encode and decode JSON data. Includes error handling for invalid JSON. * **Threading:** Uses `threading` to handle multiple clients concurrently. The `thread.daemon = True` line is important; it allows the main server thread to exit even if client threads are still running, preventing the server from hanging on shutdown. * **Data Storage:** Uses a simple `data_store` dictionary for storing data. **Important:** This is *not* persistent. If the server restarts, the data is lost. For a real-world application, you would replace this with a database (e.g., SQLite, PostgreSQL, MongoDB). * **Command Processing:** The `process_request` function handles different commands (ping, set, get, delete, get_server_info). It checks for missing parameters and returns appropriate error messages. * **Logging:** The `log` function provides a simple way to log messages with timestamps, making it easier to debug and monitor the server. * **Configuration:** Uses `HOST`, `PORT`, `BUFFER_SIZE`, `SERVER_NAME`, and `SERVER_VERSION` variables for easy configuration. * **Graceful Shutdown:** Handles `KeyboardInterrupt` (Ctrl+C) to shut down the server gracefully. * **Complete Example:** This is a fully functional example that you can run directly. * **Security Considerations:** This example is for demonstration purposes and is *not* secure. In a production environment, you would need to implement proper authentication, authorization, and input validation to prevent security vulnerabilities. Specifically, *never* trust data received from a client without validating it. * **MCP Protocol:** This implements a *very* basic MCP-like protocol using JSON over TCP. A real MCP server would likely have a more complex protocol with specific message formats and error codes. How to run: 1. **Save:** Save the code as a Python file (e.g., `mcp_server.py`). 2. **Run:** Execute the file from your terminal: `python mcp_server.py` How to test (using `netcat` or a similar tool): 1. **Open a terminal:** 2. **Connect:** `nc localhost 12345` 3. **Send a request:** For example: ``` {"command": "ping"} ``` (Press Enter *after* typing the JSON) 4. **Receive the response:** You should see: ``` {"status": "success", "message": "pong"} ``` 5. **Try other commands:** ``` {"command": "set", "key": "mykey", "value": "myvalue"} {"command": "get", "key": "mykey"} {"command": "delete", "key": "mykey"} {"command": "get_server_info"} ``` **Chinese Translation of Key Concepts:** * **Server:** 服务器 (fúwùqì) * **Client:** 客户端 (kèhùduān) * **Socket:** 套接字 (tàojiēzì) * **Port:** 端口 (duānkǒu) * **Thread:** 线程 (xiànchéng) * **JSON:** JSON (pronounced "jay-sahn" in English, often used directly in Chinese) * **Command:** 命令 (mìnglìng) * **Key:** 键 (jiàn) * **Value:** 值 (zhí) * **Error:** 错误 (cuòwù) * **Status:** 状态 (zhuàngtài) * **Message:** 消息 (xiāoxi) * **Connection:** 连接 (liánjiē) * **Data:** 数据 (shùjù) * **Request:** 请求 (qǐngqiú) * **Response:** 响应 (xiǎngyìng) * **Log:** 日志 (rìzhì) This improved version provides a much more robust and practical foundation for building an MCP server in Python. Remember to adapt it to your specific needs and security requirements.

AutoBrowser MCP

Autobrowser MCP 是一个模型上下文提供器 (MCP) 服务器，它允许 AI 应用程序控制您的浏览器。

BullMQ MCP Server

Enables AI assistants to manage BullMQ Redis-based job queues through natural language, supporting operations like job monitoring, queue control, and multi-instance Redis connections. Users can add, retry, promote, and clean jobs while accessing detailed job logs and queue statistics directly within the assistant.

Swagger MCP

An MCP server that connects to a Swagger specification and helps an AI to build all the required models to generate a MCP server for that service.

MCP Server

An implementation of the Model Context Protocol (MCP) server that enables multiple clients to connect simultaneously and handles basic context management and messaging with an extendable architecture.

Feather Code MCP Server

A GitHub integration for Claude Desktop that provides access to GitHub features directly from Claude, offering 15 powerful tools for repository management, issues, pull requests, and code operations.

otparse-mcp

A containerized MCP server for parsing OT/ICS packet captures, specifically supporting Modbus and BACnet protocols via tshark. It provides structured JSON output from PCAP files to enable LLMs to analyze industrial network traffic and derive device inventories.

mcp-llm

一个 MCP 服务器，为大型语言模型提供访问其他大型语言模型的能力。 (Or, more formally:) 一个 MCP 服务器，为 LLM 提供访问其他 LLM 的权限。

JVLink MCP Server

Enables natural language queries and analysis of Japanese horse racing data from JRA-VAN without writing SQL. Supports analyzing race results, jockey performance, breeding trends, and track conditions through conversation with Claude.

ClawHire MCP

An employer-facing MCP server that enables hiring managers to post jobs, search for candidates, and manage applications through natural language conversation. It integrates with WonderCV's hiring infrastructure and features a unique system for identifying and filtering talent with AI-agent fluency.

Obsidian MCP Tools

A read-only toolkit for searching and analyzing Markdown note directories and Obsidian vaults through AI clients. It enables metadata extraction, full-text search, and natural language querying of note content, tags, and backlinks.

Substreams Search MCP Server

MCP server that lets AI agents search the substreams.dev package registry.

HowToCook-MCP Server

An MCP server that transforms AI assistants into personal chefs by providing recipe recommendations and meal planning features based on the HowToCook repository.

MCP Mailtrap Server

官方 mailtrap.io MCP 服务器

Nanoleaf MCP Server

A Model Context Protocol server that enables controlling Nanoleaf smart lights through Warp terminal or any MCP-compatible client, providing tools for device discovery, authorization, and control of lights, brightness, colors, and effects.

Document Parser MCP

An MCP server that uses the Docling toolkit to convert various document formats, including PDFs, Office files, images, and audio, into clean Markdown for AI processing. It supports multiple processing pipelines like VLM and ASR with intelligent auto-detection and job queue management.