发现优秀的 MCP 服务器

MCP Atlassian Server

Enables AI assistants to interact with Atlassian Cloud APIs for Confluence and Jira, supporting document management, search, issue tracking, and sprint operations through natural language.

HNPX SDK

Enables AI-guided hierarchical fiction writing through structured XML documents. Supports step-by-step narrative expansion from book-level planning down to individual paragraphs, with custom Roo Code modes for collaborative story development.

Zoom Phone + Virtual Agent MCP Server

Exposes Zoom Phone call data and Zoom Virtual Agent knowledge-base/engagement data as tools for AI clients, authenticated via server-to-server OAuth.

feiniunas-mcp

MCP server for managing FeiNiu fnOS NAS via AI chat. Supports system status, Docker container operations, and disk information retrieval.

Context+

Context+ is a semantic intelligence server that transforms codebases into searchable, hierarchical feature graphs using RAG, Tree-sitter AST, and spectral clustering. It provides tools for deep code discovery, blast radius analysis, and memory graph management for high-accuracy engineering tasks.

Supabase MCP Server

A production-ready MCP server providing AI assistants with intelligent Supabase database access, featuring dynamic schema discovery, complete user management, and file storage operations.

markdown-vault-mcp

A generic Markdown vault MCP server with FTS5 full-text search, semantic vector search, frontmatter-aware indexing, incremental reindexing, and non-markdown attachment support that exposes search, read, write, and edit tools.

nucleus-apple-mcp

A macOS MCP server that provides AI agents with secure access to Calendar, Reminders, Notes, and Health data via native Apple APIs.

Desktop Window Screenshot MCP

Enables taking screenshots of the entire desktop or specific windows, with automatic clipboard integration and support for Windows, macOS, and Linux.

mcp-data-pa

Enables searching and querying Pennsylvania Open Data datasets via the Socrata API, including dataset search, SoQL queries, and metadata retrieval.

AniList MCP Server

Enables searching and browsing anime/manga, characters, staff, and more on AniList without credentials, and with a one-time login allows managing personal lists, favorites, follows, and activity.

sqlite-reader-mcp

A lightweight MCP server that provides read-only access to SQLite databases, allowing users to execute SELECT queries, list tables, and describe table schemas.

deepl-i18n

MCP server for translating JSON localization files via DeepL API or local LLMs, enabling agents to estimate, check, and run translations.

croquis-mcp

Enables AI agents to manage web production projects by providing tools for sitemap planning, content drafting, and review comment workflows.

addigy-mcp

Unified MCP server for Addigy device management, exposing both v2 and legacy v1 APIs to query devices, alerts, policies, and applications.

mcp-searcher

A template for creating MCP servers in Node.js/TypeScript with a modular architecture. It simplifies adding new tools and includes an example Elasticsearch search tool.

codex-discord-mcp

A local-first Discord bridge for Codex CLI, enabling Discord-to-Codex relay, Codex-to-Discord MCP tools, and hybrid workflows with access control and attachment safety.

permitverdict-mcp

Provides tools for verifying crypto price events and building permit intelligence, with free and paid tools using the x402 protocol.

spoonacular-mcp

Enables AI assistants to search for recipes, get nutritional information, find ingredients, and more through the Spoonacular Food API using natural language.

tokenizer-mcp

An MCP server for counting tokens in text or files using model-specific tokenizers, and reporting file metrics like lines, characters, and size.

Cocos Creator MCP Server Plugin

A comprehensive plugin that enables AI assistants to interact with the Cocos Creator editor through the Model Context Protocol, providing 80 tools across 9 categories for nearly complete editor control.

PCD

Provides a portable context layer for MCP clients, letting users load containers with structured context, work state, and accumulated knowledge across different AI environments. It enables persistent, cross-client context management through a single MCP endpoint.

YouGile Secure MCP

Provides read-only access to YouGile project management data including tasks, boards, columns, and users via the YouGile API v2.

keqing-kb

MCP server for querying the Keqing customer service knowledge base, enabling AI agents to search structured knowledge entries and retrieve authoritative, citable answers.

WhatsApp MCP Server

Enables Claude to read and send WhatsApp messages through a self-hosted REST API, supporting multi-user sessions, media, groups, contacts, and webhooks.

productive-mcp

An MCP server for Productive.io that enables users to log time, inspect projects, and manage time entries using natural language commands. It features fuzzy project matching, local caching, and remembers default services per project for streamlined time tracking.

GitHub MCP Server

MCP server that enables AI agents to perform GitHub operations like creating repositories, issues, and commits through natural language.

Security Controls MCP Server

Enables searching and bidirectional mapping of 1,451 security controls across 262 SCF-mapped frameworks, including ISO 27001, NIST CSF, DORA, and many others, through natural language queries.

Debug MCP Server in VSCode

Okay, here's a basic example of a Minecraft Protocol (MCP) server in Python, along with instructions on how to set it up for debugging in VS Code. This is a *very* simplified example and doesn't implement the full Minecraft protocol. It's designed to be a starting point for learning and debugging. **1. Install Necessary Libraries:** ```bash pip install twisted ``` **2. Create the Python Server File (e.g., `mcp_server.py`):** ```python from twisted.internet import reactor, protocol from twisted.internet.endpoints import TCP4ServerEndpoint from twisted.protocols.basic import LineReceiver class MinecraftServerProtocol(LineReceiver): delimiter = b'\n' # Use newline as the delimiter def connectionMade(self): print(f"Client connected: {self.transport.getHost()}") self.sendLine(b"Hello from the MCP server!") # Send a greeting def lineReceived(self, line): try: decoded_line = line.decode('utf-8') print(f"Received: {decoded_line}") # Simple command handling (example) if decoded_line.lower() == "ping": self.sendLine(b"pong") elif decoded_line.lower() == "time": import time self.sendLine(str(time.time()).encode('utf-8')) # Send current time else: self.sendLine(b"Unknown command") except UnicodeDecodeError: print("Received invalid UTF-8 data") self.sendLine(b"Error: Invalid input") def connectionLost(self, reason): print(f"Client disconnected: {reason}") class MinecraftServerFactory(protocol.Factory): protocol = MinecraftServerProtocol if __name__ == '__main__': endpoint = TCP4ServerEndpoint(25565, reactor) # Listen on port 25565 endpoint.listen(MinecraftServerFactory()) print("MCP server started on port 25565") reactor.run() ``` **Explanation:** * **`twisted`:** This is an asynchronous networking library. It's well-suited for handling network connections efficiently. * **`MinecraftServerProtocol`:** This class handles the communication with each client. * `connectionMade()`: Called when a client connects. Sends a greeting. * `lineReceived()`: Called when a line of data is received from the client. Decodes the line, prints it, and handles simple commands (ping, time). Includes error handling for invalid UTF-8. * `connectionLost()`: Called when a client disconnects. * **`MinecraftServerFactory`:** Creates instances of the `MinecraftServerProtocol` when a new connection is made. * **`if __name__ == '__main__':`:** This block starts the server when the script is run directly. * `TCP4ServerEndpoint`: Creates a TCP server endpoint that listens on port 25565. * `reactor.run()`: Starts the Twisted event loop, which handles all the network events. **3. Configure VS Code for Debugging:** 1. **Open the project folder in VS Code.** 2. **Create a `.vscode` folder** in the root of your project (if it doesn't exist). 3. **Create a `launch.json` file** inside the `.vscode` folder. This file tells VS Code how to run and debug your Python script. ```json { "version": "0.2.0", "configurations": [ { "name": "Python: MCP Server", "type": "python", "request": "launch", "program": "${workspaceFolder}/mcp_server.py", "console": "integratedTerminal", "justMyCode": false // Important for debugging Twisted code } ] } ``` **Explanation of `launch.json`:** * `"name"`: A descriptive name for your debug configuration. * `"type"`: Specifies that you're debugging Python code. * `"request"`: `"launch"` means you're starting the program from VS Code. * `"program"`: The path to your Python script. `${workspaceFolder}` is a VS Code variable that represents the root of your project. * `"console"`: `"integratedTerminal"` means the output will be displayed in VS Code's integrated terminal. * `"justMyCode": false`: **Crucially important for Twisted!** Twisted uses a lot of internal code. Setting this to `false` allows you to step into Twisted's code while debugging, which is often necessary to understand what's going on. **4. Run and Debug:** 1. **Set a breakpoint:** Click in the left margin of the `mcp_server.py` file next to a line of code where you want the debugger to pause (e.g., inside the `lineReceived` function). 2. **Start debugging:** Go to the "Run and Debug" view in VS Code (the icon looks like a play button with a bug). Select "Python: MCP Server" from the dropdown and click the green "Start Debugging" button (or press F5). 3. **The server will start.** You should see the "MCP server started on port 25565" message in the terminal. **5. Test with a Simple Client (netcat or similar):** Open a terminal and use `netcat` (or `nc`) to connect to the server: ```bash nc localhost 25565 ``` Type some text and press Enter. You should see the output in the VS Code terminal where the server is running, and the debugger should pause at your breakpoint. Try sending "ping" or "time". **Important Considerations and Improvements:** * **Error Handling:** The example has basic error handling, but you'll need to add more robust error handling for production code. Consider logging errors to a file. * **Protocol Implementation:** This is a *very* basic example. The real Minecraft protocol is much more complex and uses binary data. You'll need to study the Minecraft protocol documentation to implement it correctly. Libraries like `struct` in Python are essential for packing and unpacking binary data. * **Asynchronous Programming:** Twisted is an asynchronous framework. Make sure you understand how asynchronous programming works to avoid blocking the event loop. Use `defer.inlineCallbacks` and `yield` for asynchronous operations. * **Security:** If you're building a real server, security is paramount. Implement proper authentication, authorization, and input validation to prevent attacks. * **Threading/Multiprocessing:** For a high-performance server, you might need to use threads or processes to handle multiple clients concurrently. Twisted can be integrated with threads, but be careful to avoid race conditions and deadlocks. * **Logging:** Implement proper logging to track server activity and diagnose problems. **Simplified Chinese Translation of Key Phrases:** * "Client connected": 客户端已连接 (Kèhùduān yǐ liánjiē) * "Received": 收到 (Shōudào) * "Unknown command": 未知命令 (Wèi zhī mìnglìng) * "Client disconnected": 客户端已断开连接 (Kèhùduān yǐ duàn kāi liánjiē) * "MCP server started on port 25565": MCP 服务器已在端口 25565 上启动 (MCP fúwùqì yǐ zài duānkǒu 25565 shàng qǐdòng) * "Error: Invalid input": 错误：无效输入 (Cuòwù: Wúxiào shūrù) * "MCP Server": MCP 服务器 (MCP fúwùqì) * "Start Debugging": 开始调试 (Kāishǐ tiáoshì) * "Breakpoint": 断点 (Duàndiǎn) This example provides a foundation for building a more complete MCP server. Remember to consult the Minecraft protocol documentation and use appropriate libraries for handling binary data and asynchronous operations. Good luck!

Andiora Developer MCP

Enables developers to read project context, create documentation drafts, update reports, and retrieve tenant-scoped metrics through the Andiora platform APIs.