发现优秀的 MCP 服务器
通过 MCP 服务器扩展您的代理能力，拥有 86,267 个能力。
百度图片搜索 MCP Server
基于百度图片搜索API，支持图片搜索、下载及格式转换的MCP服务器。
agendum
Project memory and scoping engine for AI coding agents. It gives any agent persistent project state, bounded work packages, and cross-session continuity.
hyprland-mcp
Enables AI agents to control a Hyprland Wayland desktop by listing windows, capturing screenshots, and sending input to a dedicated agent workspace without disrupting the user's screen.
mcp-servers
待定 (dài dìng)
clevertap-mcp
An MCP server that enables AI assistants to interact with the CleverTap REST API to manage user profiles, events, campaigns, and reports. It supports multi-project configurations and provides tools for data analysis and campaign management through natural language.
Demo MCP Basic
好的，这是 MCP 服务器的 HTTP SSE 演示以及一个客户端： **服务器端 (Python - 使用 Flask):** ```python from flask import Flask, Response, request import time import json app = Flask(__name__) # 模拟 MCP 数据 def generate_mcp_data(): counter = 0 while True: data = { "timestamp": time.time(), "counter": counter, "message": f"MCP Data Update: {counter}" } yield f"data: {json.dumps(data)}\n\n" counter += 1 time.sleep(1) # 每秒更新一次 @app.route('/mcp_stream') def mcp_stream(): return Response(generate_mcp_data(), mimetype="text/event-stream") if __name__ == '__main__': app.run(debug=True, host='0.0.0.0', port=5000) ``` **代码解释:** * **`Flask`:** 使用 Flask 框架创建一个简单的 Web 服务器。 * **`generate_mcp_data()`:** 这是一个生成器函数，它模拟 MCP 数据。 * 它创建一个包含时间戳、计数器和消息的字典。 * 它使用 `yield` 关键字返回一个格式化为 SSE 事件的数据字符串。 `data: {json.dumps(data)}\n\n` 是 SSE 格式的关键。 `data:` 表示数据字段，`\n\n` 表示事件的结束。 * `time.sleep(1)` 模拟数据更新的间隔（这里是每秒一次）。 * **`/mcp_stream` 路由:** * `@app.route('/mcp_stream')` 定义了一个路由，当客户端访问 `/mcp_stream` 时，会调用 `mcp_stream()` 函数。 * `Response(generate_mcp_data(), mimetype="text/event-stream")` 创建一个 HTTP 响应，其中包含 `generate_mcp_data()` 生成的数据流，并将 `mimetype` 设置为 `text/event-stream`。 这是告诉客户端这是一个 SSE 流的关键。 * **`app.run(...)`:** 启动 Flask 服务器。 `host='0.0.0.0'` 允许从任何 IP 地址访问服务器。 **客户端 (JavaScript - HTML):** ```html <!DOCTYPE html> <html> <head> <title>MCP SSE Client</title> </head> <body> <h1>MCP Data Stream</h1> <div id="mcp-data"></div> <script> const eventSource = new EventSource('/mcp_stream'); // 替换为你的服务器地址 eventSource.onmessage = (event) => { const data = JSON.parse(event.data); const mcpDataElement = document.getElementById('mcp-data'); mcpDataElement.innerHTML += `<p>${JSON.stringify(data)}</p>`; }; eventSource.onerror = (error) => { console.error("EventSource failed:", error); eventSource.close(); // 关闭连接，防止无限重试 }; </script> </body> </html> ``` **代码解释:** * **`EventSource`:** `new EventSource('/mcp_stream')` 创建一个 `EventSource` 对象，连接到服务器的 `/mcp_stream` 端点。 确保将 `/mcp_stream` 替换为你的服务器地址（例如 `http://localhost:5000/mcp_stream`）。 * **`onmessage` 事件处理程序:** * `eventSource.onmessage = (event) => { ... }` 定义了一个事件处理程序，当服务器发送新数据时，该处理程序会被调用。 * `event.data` 包含服务器发送的数据。 * `JSON.parse(event.data)` 将 JSON 字符串解析为 JavaScript 对象。 * `document.getElementById('mcp-data')` 获取 HTML 中 `id` 为 `mcp-data` 的元素。 * `mcpDataElement.innerHTML += `<p>${JSON.stringify(data)}</p>`;` 将接收到的数据添加到 HTML 元素中。 * **`onerror` 事件处理程序:** * `eventSource.onerror = (error) => { ... }` 定义了一个事件处理程序，当发生错误时，该处理程序会被调用。 * `console.error("EventSource failed:", error);` 将错误信息输出到控制台。 * `eventSource.close();` 关闭 `EventSource` 连接，防止客户端无限重试连接。 **如何运行:** 1. **保存文件:** 将 Python 代码保存为 `mcp_server.py`，将 HTML 代码保存为 `mcp_client.html`。 2. **安装 Flask:** 在命令行中运行 `pip install flask`。 3. **运行服务器:** 在命令行中运行 `python mcp_server.py`。 4. **打开客户端:** 在浏览器中打开 `mcp_client.html`。 **预期结果:** 你将在浏览器中看到一个标题为 "MCP Data Stream" 的页面，并且页面会不断更新，显示来自服务器的 MCP 数据。 每次服务器发送新数据时，都会在页面上添加一个新的 `<p>` 元素，显示 JSON 格式的数据。 **重要注意事项:** * **CORS (跨域资源共享):** 如果你的客户端和服务器运行在不同的域名或端口上，你可能需要配置 CORS。 你可以使用 Flask 的 `flask_cors` 扩展来处理 CORS。 例如： ```python from flask import Flask, Response from flask_cors import CORS import time import json app = Flask(__name__) CORS(app) # 允许所有来源的跨域请求 # ... (其余代码不变) ``` 然后运行 `pip install flask_cors`。 * **错误处理:** 在实际应用中，你需要更完善的错误处理机制，例如处理连接错误、数据解析错误等。 * **数据格式:** 根据你的实际需求调整 MCP 数据的格式。 * **服务器地址:** 确保客户端中的 `EventSource` 连接到正确的服务器地址。 这个演示提供了一个基本的 MCP 服务器和客户端的框架。你可以根据你的具体需求进行修改和扩展。 希望这个例子能帮助你理解如何使用 HTTP SSE 实现 MCP 服务器。
Merit MCP
Connects to Merit bookkeeping software
mcp-heimdall
Security scanner for MCP servers — vet an MCP before you wire it into an agent. Detects prompt-injection, credential exfiltration (via taint analysis), RCE, and supply-chain risks, and catches cross-server exfil chains no single server reveals. Zero-dependency local CLI, SARIF output, CI-gateable, no account.
ollama-web-tools-mcp
MCP server exposing web_search, web_fetch, and cluesift_domain_fetch tools with configurable Ollama API keys and HTTP or stdio transport.
GLM Vision Server
Enables image analysis using GLM-4.5V's vision capabilities from Z.AI. Supports analyzing both local image files and URLs with customizable prompts and parameters.
CNBizAPI MCP Server
Enables AI agents to query detailed information on 77M+ Chinese companies, including basic data, shareholders, legal risks, and more, through a pay-per-query MCP server.
minio-mcp-server
Enables LLM agents to interact with MinIO/S3 object storage, supporting bucket and object operations like listing, reading, writing, and generating presigned URLs.
RGM MCP Server
Enables Revenue Growth Management analysis for the Alco-Bev industry using NIQ data, including price elasticity, promo effectiveness, and pricing optimization.
PubCrawl
PubCrawl provides LLMs with access to PubMed, FDA/UK drug labeling, and ClinicalTrials.gov. It enables searching literature, retrieving abstracts and full texts, comparing US and UK drug labels, and exploring clinical trials.
Global MCP Manager
Enables executing terminal commands and managing files across different contexts: local system, remote SSH servers, and GitHub repositories. Provides comprehensive file operations, directory navigation, and multi-environment command execution capabilities.
errorbar
Enables MCP clients to call the platform's full API as individual tools—covering evals, criteria, gates, logs, datasets, aliases, label sets, audit/proving, dedicated endpoints, and training—with validated inputs and raw API responses.
Movie Booking Assistant
Provides a suite of tools for searching movies, checking showtimes, and managing ticket bookings for Bangalore cinemas. It enables AI clients to handle end-to-end movie theater interactions including seat availability checks and reservation management.
jira-lite-mcp
A lightweight Jira Cloud MCP server for Claude Code that provides high-value tools for reading and writing Jira issues with human-readable responses.
ida-pro-mcp-plus
An enhanced MCP server for IDA Pro that integrates bulk binary export, multi-instance management via Broker mode, and 80+ analysis tools for AI-assisted reverse engineering.
apple-mail-mcp
Enables AI assistants to read, send, search, and manage emails in Apple Mail on macOS.
🚀 deco.host — Instant MCP Server Deployment
开源 MCP 服务器平台。一个用于构建自定义 MCP 服务器并在任何地方部署的 Web Ninite。
Office MCP Server
Connects Claude with Microsoft 365 services such as Email, Calendar, Teams, OneDrive, and more through the Microsoft Graph API.
Zotero Chunk RAG
Enables passage-level semantic search over a Zotero library by extracting, chunking, and embedding PDF text using Gemini and ChromaDB. It provides MCP tools to perform topical searches and retrieve specific document passages with surrounding context.
Precogly MCP
MCP server for Precogly threat modeling, enabling users to integrate threat modeling capabilities into AI assistants.
mcp-python-exec-sandbox
Sandboxed Python execution with automatic dependency management. Executes Python scripts in isolated environments (bubblewrap or Docker) with PEP 723 inline dependencies, preventing host pollution.
FastMCP Production-Ready Server
A production-ready MCP server that enables users to interact with Neo4j databases through health checks and Cypher query tools. It features a structured, containerized architecture with built-in support for Azure deployments and environment-driven configuration.
Mnemonica Strategy
An MCP server that connects to running Node.js applications via the Chrome Debug Protocol to analyze Mnemonica type hierarchies at runtime. It enables users to validate and improve static analysis by comparing runtime types with Tactica-generated types.
codex-buddy-for-claude
Enables Claude Code to leverage OpenAI models for expert code review, deep architecture analysis, and security audits, with automatic markdown report generation.
Pokemon MCP Server
An MCP server that provides standardized access to Pokemon data, allowing users to search, compare, and retrieve detailed information about Pokemon through natural language tools.
ssh-mcp-dynamic
A minimal MCP server that lets an MCP client run shell commands on remote hosts over SSH, with host, key, user, and port chosen per call.