发现优秀的 MCP 服务器

Docs MCP

An MCP server that searches official documentation for popular libraries, scrapes pages, and returns clean, LLM-ready text.

Browser Automation MCP Server

一个模型上下文协议（MCP）服务器，为 Claude 和其他 MCP 兼容的 AI 助手提供浏览器自动化功能。

admob-manager-mcp

Provides AI agents with complete control over Google AdMob accounts, apps, ad units, mediation groups, A/B experiments, and performance reporting via the AdMob REST API.

Quickbase MCP Server

Provides a standardized interface for interacting with Quickbase's JSON RESTful API through Claude and other MCP clients, supporting operations like querying records, managing table relationships, and handling file attachments.

🧠 Model Context Protocol (MCP)

好的，这是使用 Langchain MCP Adapters 和 Ollama 实现 MCP 的演示： **标题：使用 Langchain MCP Adapters 和 Ollama 实现 MCP 的演示** **简介：** 本演示展示了如何使用 Langchain MCP Adapters 和 Ollama 来实现 MCP（多通道处理）。MCP 是一种技术，它允许您使用多个通道（例如，文本、图像、音频）来处理数据。这可以提高准确性和效率。 **先决条件：** * 已安装 Python 3.7 或更高版本 * 已安装 Langchain * 已安装 Ollama * 已安装 Langchain MCP Adapters **安装：** ```bash pip install langchain pip install ollama pip install langchain-mcp-adapters ``` **代码：** ```python from langchain_mcp_adapters import MultiChannelProcessor from langchain.llms import Ollama from langchain.chains import LLMChain from langchain.prompts import PromptTemplate # 1. 定义你的通道。 # 在这个例子中，我们将使用两个通道：文本和图像。 channels = [ { "name": "text", "type": "text", }, { "name": "image", "type": "image", }, ] # 2. 创建一个多通道处理器。 mcp = MultiChannelProcessor(channels=channels) # 3. 定义你的提示模板。 # 这个提示模板将用于处理文本通道。 text_prompt_template = """ 你是一个有用的助手。 请回答以下问题： {question} """ text_prompt = PromptTemplate( input_variables=["question"], template=text_prompt_template, ) # 4. 定义你的语言模型。 # 在这个例子中，我们将使用 Ollama。 llm = Ollama(model="llama2") # 5. 创建一个 LLM 链。 text_chain = LLMChain(llm=llm, prompt=text_prompt) # 6. 将 LLM 链添加到多通道处理器。 mcp.add_channel("text", text_chain) # 7. 定义一个函数来处理图像通道。 def image_processor(image_path): """ 这个函数将处理图像通道。 Args: image_path: 图像的路径。 Returns: 图像的描述。 """ # 在这里，你可以使用任何图像处理库来处理图像。 # 在这个例子中，我们将使用 PIL 库。 from PIL import Image import pytesseract # 打开图像。 image = Image.open(image_path) # 使用 pytesseract 来提取图像中的文本。 text = pytesseract.image_to_string(image) # 返回图像的描述。 return text # 8. 将图像处理器添加到多通道处理器。 mcp.add_channel("image", image_processor) # 9. 创建一个输入。 input_data = { "text": { "question": "什么是 Langchain？", }, "image": { "image_path": "image.png", }, } # 10. 处理输入。 output = mcp.process(input_data) # 11. 打印输出。 print(output) ``` **解释：** 1. **定义通道：** 首先，我们定义了两个通道：文本和图像。每个通道都有一个名称和一个类型。 2. **创建多通道处理器：** 接下来，我们创建了一个多通道处理器。多通道处理器负责处理来自不同通道的数据。 3. **定义提示模板：** 我们定义了一个提示模板，用于处理文本通道。提示模板是一个字符串，它包含一个或多个占位符。占位符将被输入数据替换。 4. **定义语言模型：** 我们定义了一个语言模型。语言模型是一个可以生成文本的模型。在这个例子中，我们使用 Ollama。 5. **创建 LLM 链：** 我们创建了一个 LLM 链。LLM 链是一个将提示模板和语言模型连接在一起的链。 6. **将 LLM 链添加到多通道处理器：** 我们将 LLM 链添加到多通道处理器。这告诉多通道处理器使用 LLM 链来处理文本通道。 7. **定义图像处理器：** 我们定义了一个图像处理器。图像处理器是一个可以处理图像的函数。 8. **将图像处理器添加到多通道处理器：** 我们将图像处理器添加到多通道处理器。这告诉多通道处理器使用图像处理器来处理图像通道。 9. **创建输入：** 我们创建了一个输入。输入是一个字典，它包含每个通道的数据。 10. **处理输入：** 我们处理输入。多通道处理器将使用 LLM 链和图像处理器来处理输入数据。 11. **打印输出：** 我们打印输出。输出是一个字典，它包含每个通道的处理结果。 **结论：** 本演示展示了如何使用 Langchain MCP Adapters 和 Ollama 来实现 MCP。MCP 是一种强大的技术，它可以提高准确性和效率。 **注意：** * 您需要将 `image.png` 替换为实际的图像文件。 * 您需要根据您的需要修改提示模板和图像处理器。 * 您可以使用任何语言模型来代替 Ollama。 希望这个演示对您有所帮助！

USPTO Patent Citation MCP Server

Provides access to USPTO enriched citation and office action citation APIs with smart context reduction and progressive disclosure workflows for patent citation analysis.

instantly-ai-mcp

Enables AI assistants to interact with the Instantly.ai v2 API to manage campaigns, leads, accounts, replies, blocklist, and webhooks, with tools organized into read, write, and dangerous safety tiers.

tourism-mcp-server

Enables searching for national-standard scenic areas, rural tourism villages, accommodations (including restaurants, guesthouses, and starred hotels), and outbound travel agencies, with filtering by region, category, and business status.

Garmin Connect MCP Server

Connects Garmin Connect data to MCP-compatible clients, providing access to fitness activities, health metrics, and training plans. It supports advanced features like headless 2FA and automated MFA retrieval to enable seamless health data interaction through natural language.

gdrive-mcp

Enables MCP clients to access a configured Google Drive folder over Streamable HTTP, listing files and reading text content with automatic export of Google Docs, Sheets, and Slides.

Wealthsimple Help Center MCP

Enables AI agents to search, browse, and retrieve articles from the Wealthsimple Help Center via 7 typed tools (search, taxonomy, article retrieval) using the public Zendesk API.

Super Subagents

Spawn parallel autonomous AI agent sessions from a single MCP client. Each agent gets its own workspace, tools, and execution context.

docs-mcp

Upload Word, Excel, PDF, or PowerPoint documents to a vector RAG store with vision-model extraction, then search semantically and retrieve chunks with page numbers for precise citations.

Mallory MCP Server

A robust Model Control Protocol server that enables AI agents to access real-time cyber threat intelligence and detailed information about vulnerabilities, threat actors, malware, and other cyber-security entities.

Weather MCP Tool

A Model Context Protocol tool that provides weather information for cities, with London access requiring Solana devnet payment via the Latinum Wallet MCP server.

pymdu

Enables Claude Desktop or Gemini CLI to communicate with PyMDU, allowing analysis and processing of urban data using PyMDU tools through natural language commands.

Polymath MCP

Aggregates 17 free, keyless research sources into a single MCP server, enabling unified search for papers, code, models, trends, standards, and electronic components with deduplicated results.

mcp-arcgis-sanjuannm

Enables querying and searching San Juan County, New Mexico open geospatial data (parcels, addresses, zoning, public works) via ArcGIS feature services.

SOAP MCP Server

Exposes SOAP web services as Model Context Protocol (MCP) servers, allowing AI models to interact with legacy SOAP services through automatic method discovery and type mapping.

MCP Servers

一个 Node.js 和 TypeScript 服务器项目，提供了一个简单的入门示例，使用 Express.js Web 服务器，并支持热重载、测试和模块化结构。

hypersignal

Monetized MCP server for Hyperliquid whale/funding/risk analytics + signals;

MCP Jibun Server

Enables AI agents to read and retrieve the latest posts from Jibun and Ech0 instances. It allows users to list configured sources and fetch content with support for pagination and source selection.

cloro MCP Server

MCP server that gives agents tools to query ChatGPT, Perplexity, Gemini, Copilot, Grok, Google AI Mode, Google Search, and Google News through the cloro API, returning parsed answers with cited sources.

Secure MCP Server Template

A template for creating secure, remotely accessible MCP servers with OAuth authentication and Cloudflare ZeroTrust Access protection. Enables deployment of containerized MCP servers that can be safely accessed by Claude Desktop and other MCP clients through authenticated connections.

MCP Healthcare System

A production-grade MCP server that enables AI assistants to securely manage healthcare data through clinical tools for patient vitals, lab results, and medication ordering. It prioritizes security and compliance with features like HIPAA-ready audit logging, PII redaction, and role-based access control.

nexus-adf

Enables Claude to diagnose and fix Azure Data Factory pipeline failures with automated rollback.

izan.io

Turns your Chrome browser into an MCP server, allowing AI clients to control browser actions like clicking, typing, navigating, and data extraction through custom JavaScript tools.

DocuQueue MCP Server

Enables creating professional documents (invoices, contracts, certificates, proposals, reports) via the DocuQueue API, with tools for template management, filling, previewing, and PDF generation.