Discover Awesome MCP Servers

web-search

Enables web searching using Google search results with no API keys required.

predictfun-mcp

MCP (Model Context Protocol) server that gives AI agents structured access to Predict.fun — a prediction market protocol on BNB Chain with $1.5B+ volume and yield-bearing mechanics via Venus Protocol. Indexes data from three subgraphs: orderbook activity, position lifecycle, and yield mechanics.

Ring MCP

An MCP server for Claude Code that shows always-on-top notification popups with sound when tasks complete, with OS-native notification fallback.

MCP Server Implementation Guide

以下是一个指南和实现，用于创建你自己的 MCP (模型控制协议) 服务器，以便与 Cursor 集成： **标题：创建你自己的 Cursor 集成 MCP 服务器指南与实现** **简介：** Cursor 是一款强大的代码编辑器，它允许通过 MCP (Model Control Protocol) 与外部语言模型进行交互。 本指南将引导你完成创建自己的 MCP 服务器的过程，以便将你自己的语言模型集成到 Cursor 中。 **1. 了解 MCP (Model Control Protocol):** * **目的：** MCP 是一种允许 Cursor 与外部语言模型进行通信的协议。 它定义了 Cursor 如何向模型发送请求以及模型如何返回响应。 * **通信方式：** MCP 通常使用 JSON over WebSocket 进行通信。 * **关键消息类型：** * **`completion` 请求：** Cursor 向模型发送代码补全请求。 * **`completion` 响应：** 模型返回代码补全建议。 * **`chat` 请求：** Cursor 向模型发送聊天请求。 * **`chat` 响应：** 模型返回聊天回复。 * **`edit` 请求：** Cursor 向模型发送代码编辑请求。 * **`edit` 响应：** 模型返回代码编辑建议。 * **`health` 请求：** Cursor 向服务器发送健康检查请求。 * **`health` 响应：** 服务器返回健康状态。 **2. 选择编程语言和框架：** 你可以使用任何你喜欢的编程语言和框架来构建 MCP 服务器。 一些常见的选择包括： * **Python:** 使用 `websockets` 或 `aiohttp` 库。 * **Node.js:** 使用 `ws` 或 `socket.io` 库。 * **Go:** 使用 `gorilla/websocket` 库。 本指南将使用 Python 和 `websockets` 库作为示例。 **3. 设置 WebSocket 服务器：** 首先，你需要设置一个 WebSocket 服务器来监听来自 Cursor 的连接。 ```python import asyncio import websockets import json async def handle_connection(websocket, path): print(f"New connection from {websocket.remote_address}") try: async for message in websocket: print(f"Received message: {message}") try: data = json.loads(message) # 处理消息 response = await process_message(data) await websocket.send(json.dumps(response)) except json.JSONDecodeError: print("Invalid JSON received") await websocket.send(json.dumps({"error": "Invalid JSON"})) except Exception as e: print(f"Error processing message: {e}") await websocket.send(json.dumps({"error": str(e)})) except websockets.exceptions.ConnectionClosedError: print(f"Connection closed unexpectedly from {websocket.remote_address}") except websockets.exceptions.ConnectionClosedOK: print(f"Connection closed normally from {websocket.remote_address}") finally: print(f"Connection closed from {websocket.remote_address}") async def process_message(data): # 在这里处理不同类型的 MCP 请求 if data.get("type") == "completion": return await handle_completion(data) elif data.get("type") == "chat": return await handle_chat(data) elif data.get("type") == "edit": return await handle_edit(data) elif data.get("type") == "health": return await handle_health(data) else: return {"error": "Unknown message type"} async def handle_completion(data): # TODO: 调用你的语言模型进行代码补全 prompt = data.get("prompt") # 示例：返回一个简单的补全建议 completion = f"// This is a completion for: {prompt}" return {"completion": completion} async def handle_chat(data): # TODO: 调用你的语言模型进行聊天 message = data.get("message") # 示例：返回一个简单的聊天回复 response = f"You said: {message}" return {"response": response} async def handle_edit(data): # TODO: 调用你的语言模型进行代码编辑 code = data.get("code") instruction = data.get("instruction") # 示例：返回一个简单的编辑建议 edited_code = f"// Edited code based on: {instruction}\n{code}" return {"edited_code": edited_code} async def handle_health(data): # 返回服务器的健康状态 return {"status": "ok"} async def main(): async with websockets.serve(handle_connection, "localhost", 8765): print("WebSocket server started at ws://localhost:8765") await asyncio.Future() # 保持服务器运行 if __name__ == "__main__": asyncio.run(main()) ``` **4. 处理 MCP 请求：** 在 `process_message` 函数中，你需要根据 `data.get("type")` 的值来处理不同类型的 MCP 请求。 * **`completion` 请求：** * 从 `data` 中提取代码补全所需的上下文信息（例如，当前代码、光标位置等）。 * 调用你的语言模型来生成代码补全建议。 * 将补全建议封装在 `completion` 响应中并返回。 * **`chat` 请求：** * 从 `data` 中提取聊天消息。 * 调用你的语言模型来生成聊天回复。 * 将回复封装在 `chat` 响应中并返回。 * **`edit` 请求：** * 从 `data` 中提取代码和编辑指令。 * 调用你的语言模型来生成代码编辑建议。 * 将编辑后的代码封装在 `edit` 响应中并返回。 * **`health` 请求：** * 返回服务器的健康状态。 **5. 集成你的语言模型：** 在 `handle_completion`、`handle_chat` 和 `handle_edit` 函数中，你需要集成你自己的语言模型。 这可能涉及： * 加载你的语言模型。 * 预处理输入数据。 * 调用语言模型进行推理。 * 后处理输出数据。 **6. 配置 Cursor：** 1. 打开 Cursor 的设置。 2. 搜索 "Model Control Protocol"。 3. 启用 "Enable Model Control Protocol"。 4. 在 "Model Control Protocol URL" 中输入你的 MCP 服务器的 URL (例如，`ws://localhost:8765`)。 **7. 测试：** 1. 运行你的 MCP 服务器。 2. 在 Cursor 中打开一个代码文件。 3. 尝试代码补全、聊天或代码编辑功能。 4. 检查你的 MCP 服务器是否收到请求并返回了正确的响应。 **8. 错误处理：** * 在服务器端，捕获所有可能的异常并返回包含错误信息的 JSON 响应。 * 在 Cursor 端，检查响应中是否包含错误信息并向用户显示。 **9. 优化：** * **性能：** 优化你的语言模型和 MCP 服务器以提高性能。 * **可扩展性：** 设计你的 MCP 服务器以支持多个并发连接。 * **安全性：** 考虑安全性问题，例如身份验证和授权。 **示例 JSON 消息格式：** **Completion Request:** ```json { "type": "completion", "prompt": "def hello_world():\n " } ``` **Completion Response:** ```json { "completion": "print('Hello, world!')" } ``` **Chat Request:** ```json { "type": "chat", "message": "How do I write a for loop in Python?" } ``` **Chat Response:** ```json { "response": "You can write a for loop in Python like this: `for i in range(10): print(i)`" } ``` **Edit Request:** ```json { "type": "edit", "code": "def add(a, b):\n return a + b", "instruction": "Add a docstring to the function." } ``` **Edit Response:** ```json { "edited_code": "def add(a, b):\n \"\"\"Adds two numbers together.\"\"\"\n return a + b" } ``` **Health Request:** ```json { "type": "health" } ``` **Health Response:** ```json { "status": "ok" } ``` **总结：** 通过遵循本指南，你可以创建自己的 MCP 服务器，并将你自己的语言模型集成到 Cursor 中。 这将使你能够利用你自己的模型来增强 Cursor 的代码补全、聊天和代码编辑功能。 记住，这只是一个起点，你需要根据你的具体需求进行调整和优化。 **重要提示：** * 确保你的语言模型符合 Cursor 的使用条款和隐私政策。 * 仔细测试你的 MCP 服务器，以确保其稳定性和可靠性。 * 考虑安全性问题，例如身份验证和授权。 This translation provides a comprehensive guide and implementation example for creating your own MCP server for Cursor integration. It covers the key concepts, steps, and considerations involved in the process. Remember to replace the placeholder comments with your actual language model integration logic. Good luck!

ultra-confluence-mcp

A Model Context Protocol server for Confluence Cloud that aggressively trims API responses to reduce token usage, with features like disk offload and tunable tool surface.

EDINET DB MCP Server

Structured financial data for ~3,800 Japanese listed companies from EDINET regulatory filings — financials, major shareholders, segments, executive compensation, and corporate history. Remote MCP over HTTPS with OAuth 2.0, free tier.

My Awesome MCP

A basic MCP server built with FastMCP framework that provides example tools including message echoing and server information retrieval. Supports both stdio and HTTP transports with Docker deployment capabilities.

Diffblue Cover MCP Server

Enables AI development environments to call the Diffblue Cover CLI tool to automatically generate unit tests for Java code.

Kintone Book Management MCP Tool

A Model Context Protocol (MCP) server that provides a tool for retrieving and managing book information from a Kintone database application.

AllFlyghts MCP Server

Enables flight search, location lookup, and city information retrieval using the AllFlyghts public API through MCP tools.

Weather-server MCP Server

A TypeScript-based MCP server that provides weather information through resources and tools, allowing users to access current weather data and forecast predictions for different cities.

VeriSom

Scores contract safety for Somnia Agents using on-chain analysis, RAG, and a VeriSom contract, returning a pre-interaction safety score.

TaskFlow MCP

A task management server that helps AI assistants break down user requests into manageable tasks and track their completion with user approval steps.

Codex MCP Telegram

Enables remote execution of Codex CLI commands and provides an MCP tool for AI agents to escalate questions to humans via Telegram, allowing for human-in-the-loop workflows when away from the machine.

Deobfuscate MCP Server

An LLM-optimized server for reverse-engineering and navigating minified JavaScript bundles by splitting them into searchable, individual modules. It enables LLMs to analyze code architecture, extract specific symbols, and perform semantic searches while efficiently managing context window usage.

FinOpsGuard

MCP agent providing cost-aware guardrails for IaC in CI/CD with advanced policy enforcement.

celp-mcp

Celp-MCP enables natural-language analytics over SQL, MongoDB, and Databricks warehouses through MCP-compatible clients, converting questions into multi-step database plans and returning markdown reports.

email-mcp

MCP server for managing email via IMAP/SMTP, supporting multiple accounts and tools for reading, sending, searching, and organizing emails.

code-mode-alchemy-mcp

Gives Claude access to Alchemy blockchain APIs by loading all API specs at startup and providing two tools: search and execute. Enables querying NFTs, token balances, transactions, and more via natural language.

IB Async MCP Server

Enables read-only access to Interactive Brokers data including contracts, market data, news, fundamentals, and portfolio/account information for LLM workflows and autonomous agents.

mcp-sii

Open-source MCP server for Chile's SII free invoicing system, enabling AI agents to query issued and received tax documents.

domainkits-mcp

DomainKits MCP connects AI assistants (Claude, GPT, etc.) to professional domain intelligence tools. Instead of guessing domain availability, the AI can actually check it. Instead of generic naming suggestions, it validates ideas against real market data.

Shopify MCP Server

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

Labradoc MCP Server

Enables AI assistants to interact with Labradoc's document management, email ingestion, task extraction, and integration features through MCP tools.

Firefly III MCP Server

Enables interaction with Firefly III personal finance tools through natural language, deployed globally on Cloudflare Workers for low latency and high availability.

CPersona

Persistent AI memory server with 3-layer hybrid search (vector + FTS5 + keyword), confidence scoring via Reciprocal Rank Fusion, episodic/profile memory, and 16 tools. Zero LLM dependency. Works standalone with Claude Desktop and Claude Code. MIT licensed.

MusicMCP.AI

Enables AI-powered music generation through natural language commands, supporting both inspiration mode (AI-generated lyrics and style) and custom mode (user-provided lyrics and parameters) to create songs with direct download links.

llm-battle-mcp

Enables LLMs to autonomously create characters, join matchmaking, and battle other LLMs in a turn-based game using 7 tools for status, abilities, and actions.

MCP OpenAPI Connector

Enables Claude Desktop and other MCP clients to interact with any OAuth2-authenticated OpenAPI-based API through automatic tool generation from OpenAPI specifications, with built-in token management and authentication handling.

FlashCardMCP

Converts JSON/CSV Markdown content into interactive flashcard pages with multiple templates, PDF export, and voice support for language learning.