发现优秀的 MCP 服务器

Synology Download Station MCP Server

A Model Context Protocol server that enables AI assistants to manage downloads, search for torrents, and monitor download statistics on a Synology NAS.

Local FAISS MCP Server

Provides local vector database functionality using FAISS for document ingestion, semantic search, and Retrieval-Augmented Generation (RAG) applications with persistent storage and customizable embedding models.

python-docs-server MCP Server

镜子 (jìng zi)

arXiv Discovery MCP

MCP server for discovering, triaging, and monitoring arXiv papers with transparent interest modeling and inspectable ranking.

ministic-fishstick

Minimal high-performance MCP server for semantic code indexing and vector search using Bun, SQLite, and Tree-Sitter. It enables AI agents to index, search, and manage codebases via tools like code_index_search and code_index_start.

Swagger to MCP

Automatically converts Swagger/OpenAPI specifications into dynamic MCP tools, enabling interaction with any REST API through natural language by loading specs from local files or URLs.

MCP Server - Placeholder Implementation

An MCP server implementation in Python with placeholder tools, deployable to Azure Web App via GitHub Actions. Supports STDIO, HTTP REST, and WebSocket interfaces.

metabase-mcp

Connects AI assistants to Metabase for database queries, SQL execution, dashboard management, and analytics workflows via natural language.

Multi-Capability Proxy Server

A Flask-based server that hosts multiple tools, each exposing functionalities by calling external REST APIs through a unified interface.

MCP demo (DeepSeek as Client's LLM)

Okay, I can help you outline the steps to run a minimal client-server demo using the DeepSeek API, focusing on the core concepts and providing example code snippets. Since I can't directly execute code or set up environments, I'll give you the instructions and code you'll need to adapt and run yourself. **Important Considerations Before You Start:** * **DeepSeek API Key:** You'll need a valid DeepSeek API key. Obtain one from the DeepSeek AI platform. Keep it secure and don't hardcode it directly into your scripts (use environment variables or configuration files). * **Python Environment:** I'll assume you're using Python. Make sure you have Python 3.7+ installed. * **Libraries:** You'll need the `requests` library for making HTTP requests to the DeepSeek API. Install it using `pip install requests`. You might also want `Flask` or `FastAPI` for a simple server. **Conceptual Overview** 1. **Client:** The client sends a request to the server. In this case, the request will contain a prompt that you want DeepSeek to complete. 2. **Server:** The server receives the request from the client, calls the DeepSeek API with the prompt, gets the response from DeepSeek, and sends the response back to the client. 3. **DeepSeek API:** This is the external service that performs the language model inference. **Step-by-Step Instructions and Code Examples** **1. Server (using Flask)** ```python # server.py from flask import Flask, request, jsonify import requests import os app = Flask(__name__) # Replace with your actual DeepSeek API key (ideally from an environment variable) DEEPSEEK_API_KEY = os.environ.get("DEEPSEEK_API_KEY") # Get from environment DEEPSEEK_API_URL = "https://api.deepseek.com/v1/chat/completions" # Replace if different @app.route('/generate', methods=['POST']) def generate_text(): try: data = request.get_json() prompt = data.get('prompt') if not prompt: return jsonify({'error': 'Prompt is required'}), 400 headers = { 'Content-Type': 'application/json', 'Authorization': f'Bearer {DEEPSEEK_API_KEY}' } payload = { "model": "deepseek-chat", # Or another DeepSeek model "messages": [{"role": "user", "content": prompt}], "max_tokens": 200, # Adjust as needed "temperature": 0.7 # Adjust as needed } response = requests.post(DEEPSEEK_API_URL, headers=headers, json=payload) response.raise_for_status() # Raise HTTPError for bad responses (4xx or 5xx) deepseek_data = response.json() generated_text = deepseek_data['choices'][0]['message']['content'] return jsonify({'generated_text': generated_text}) except requests.exceptions.RequestException as e: print(f"API Request Error: {e}") return jsonify({'error': f'API Request Error: {e}'}), 500 except Exception as e: print(f"Server Error: {e}") return jsonify({'error': f'Server Error: {e}'}), 500 if __name__ == '__main__': app.run(debug=True, port=5000) # Or any port you prefer ``` **Explanation of `server.py`:** * **Imports:** Imports necessary libraries (Flask, requests, json, os). * **API Key:** Retrieves the DeepSeek API key from an environment variable. **Never hardcode your API key directly in the script!** * **Flask App:** Creates a Flask web application. * **`/generate` Route:** Defines a route that listens for POST requests at `/generate`. * **Request Handling:** * Extracts the `prompt` from the JSON request body. * Constructs the headers for the DeepSeek API request, including the `Authorization` header with your API key. * Creates the payload (JSON data) for the DeepSeek API request. This includes the model name, the prompt (formatted as a message), and other parameters like `max_tokens` and `temperature`. * Sends the request to the DeepSeek API using `requests.post()`. * Handles potential errors (e.g., network issues, invalid API key). * **Response Handling:** * Parses the JSON response from the DeepSeek API. * Extracts the generated text from the response. The exact structure of the response depends on the DeepSeek API. The code assumes a structure like `deepseek_data['choices'][0]['message']['content']`. **You might need to adjust this based on the actual DeepSeek API response format.** * Returns the generated text as a JSON response to the client. * **Error Handling:** Includes `try...except` blocks to catch potential errors during the API request and server processing. Returns error messages to the client. * **Running the App:** Starts the Flask development server. **2. Client (using Python)** ```python # client.py import requests import json SERVER_URL = "http://localhost:5000/generate" # Adjust if your server is running on a different address/port def generate_text(prompt): try: payload = {'prompt': prompt} headers = {'Content-Type': 'application/json'} response = requests.post(SERVER_URL, headers=headers, data=json.dumps(payload)) response.raise_for_status() # Raise HTTPError for bad responses (4xx or 5xx) data = response.json() generated_text = data.get('generated_text') return generated_text except requests.exceptions.RequestException as e: print(f"Request Error: {e}") return None except Exception as e: print(f"Error: {e}") return None if __name__ == '__main__': user_prompt = "Write a short story about a cat who goes on an adventure." generated_text = generate_text(user_prompt) if generated_text: print("Generated Text:") print(generated_text) else: print("Failed to generate text.") ``` **Explanation of `client.py`:** * **Imports:** Imports the `requests` and `json` libraries. * **`SERVER_URL`:** Defines the URL of the server's `/generate` endpoint. Make sure this matches the address and port where your server is running. * **`generate_text(prompt)` Function:** * Takes a `prompt` as input. * Constructs the payload (JSON data) to send to the server. * Sets the `Content-Type` header to `application/json`. * Sends a POST request to the server using `requests.post()`. * Handles potential errors (e.g., network issues, server not available). * Parses the JSON response from the server. * Extracts the `generated_text` from the response. * Returns the generated text. * **Main Execution Block:** * Sets a sample `user_prompt`. * Calls the `generate_text()` function to get the generated text. * Prints the generated text to the console. **3. Running the Demo** 1. **Set the API Key:** Before running anything, set the `DEEPSEEK_API_KEY` environment variable. How you do this depends on your operating system: * **Linux/macOS:** ```bash export DEEPSEEK_API_KEY="YOUR_DEEPSEEK_API_KEY" ``` * **Windows (Command Prompt):** ```cmd set DEEPSEEK_API_KEY=YOUR_DEEPSEEK_API_KEY ``` * **Windows (PowerShell):** ```powershell $env:DEEPSEEK_API_KEY="YOUR_DEEPSEEK_API_KEY" ``` **Replace `YOUR_DEEPSEEK_API_KEY` with your actual API key.** 2. **Run the Server:** Open a terminal or command prompt, navigate to the directory where you saved `server.py`, and run: ```bash python server.py ``` The Flask development server will start, and you'll see output indicating that it's running. 3. **Run the Client:** Open another terminal or command prompt, navigate to the directory where you saved `client.py`, and run: ```bash python client.py ``` The client will send a request to the server, the server will call the DeepSeek API, and the generated text will be printed to the client's console. **Important Notes and Troubleshooting** * **API Key:** Double-check that your API key is correct and that you've set the environment variable properly. An incorrect API key will result in an authentication error. * **Network Connectivity:** Make sure your server has internet access to reach the DeepSeek API. * **Error Messages:** Carefully examine any error messages you receive. They often provide clues about what's going wrong. * **DeepSeek API Response Format:** The code assumes a specific format for the DeepSeek API response. If the API changes its response format, you'll need to update the code accordingly. Refer to the DeepSeek API documentation for the correct format. * **Rate Limits:** Be aware of the DeepSeek API's rate limits. If you send too many requests in a short period, you might get rate-limited. Implement error handling and potentially retry logic to deal with rate limits. * **Security:** For production environments, use a more robust web server (like Gunicorn or uWSGI) instead of the Flask development server. Also, consider using HTTPS for secure communication between the client and server. * **Model Selection:** The code uses `"deepseek-chat"` as the model. Check the DeepSeek API documentation for other available models and their capabilities. * **Prompt Engineering:** The quality of the generated text depends heavily on the prompt you provide. Experiment with different prompts to get the best results. **Simplified Chinese Translation of Key Phrases** Here are some key phrases translated into Simplified Chinese: * **Prompt:** 提示 (tíshì) * **Generated Text:** 生成的文本 (shēngchéng de wénběn) * **API Key:** API 密钥 (API mìyào) * **Server:** 服务器 (fúwùqì) * **Client:** 客户端 (kèhùduān) * **Error:** 错误 (cuòwù) * **Request:** 请求 (qǐngqiú) * **Response:** 响应 (xiǎngyìng) * **Authentication:** 身份验证 (shēnfèn yànzhèng) * **Rate Limit:** 速率限制 (sùlǜ xiànzhì) This detailed guide should help you get started with a basic DeepSeek API client-server demo. Remember to adapt the code to your specific needs and consult the DeepSeek API documentation for the most up-to-date information. Good luck!

Weather MCP Server

Enables asking natural language weather questions; an LLM agent selects the appropriate live weather API tool and returns an HTML-formatted answer.

Tanda Workforce MCP Server

Integrates Tanda Workforce API with AI assistants to manage employee schedules, timesheets, leave requests, clock in/out operations, and workforce analytics through natural language with OAuth2 authentication.

claude-peers

Enables discovery and instant communication between multiple local Claude Code instances running across different projects. It allows agents to list active peers, share work summaries, and send messages through a local broker daemon.

MCPizza

An MCP server that allows AI assistants to order Domino's Pizza through an unofficial API, with features for store location, menu browsing, and order management.

blender-mcp-2.79

Enables Claude AI to control Blender 2.79 for 3D modeling, scene creation, and manipulation through the Model Context Protocol, with legacy compatibility for older Blender versions.

Semantic Scholar MCP Server

Semantic Scholar API, providing comprehensive access to academic paper data, author information, and citation networks.

ltchiptool-mcp

Enables chip-level firmware extraction and analysis for BK7231 family IoT devices via UART, supporting flash dumping, decryption, and partition extraction.

MCP-123

A lightweight implementation that provides the simplest way to set up an MCP server and client, requiring just two lines of code to create a fully functional system.

remote-mcp-server-authless

This is a stateless remote MCP server deployed on Cloudflare Workers without authentication, allowing users to register custom tools and connect to MCP clients such as Cloudflare AI Playground or Claude Desktop.

uv-mcp-server

Okay, I understand. Please provide the English text you would like me to translate to Chinese. I will focus on providing a stable and accurate translation, avoiding any "hallucinations" or nonsensical outputs.

Terminal SSH MCP Server

MCP server that enables SSH access and terminal commands for Railway containers and local shell execution.

Valjs

一个用于 Valtown 封装器的 MCP 服务器 (糟糕的描述)

mcp-ip-api

Provides IP geolocation lookups via ip-api.com, including single and batch queries up to 100 IPs.

spar-agent

Enables developers to compare their code predictions against AI-generated implementations, log misconceptions, and receive spaced-repetition learning feedback to guide their understanding of gaps.

idnow-mcp-server

An MCP server that lets LLM agents operate the IDnow Trust Platform sandbox to list verification flows, create sessions, check sessions, and get session details.

mr-model MCP Server

Standalone Model Context Protocol (MCP) server for the mr-model platform, offering 5 tools to query video lists, comments, OCR transcripts, search, and blogger opinions via a web API.

@pipeworx/congressional-documents

Full-text search and retrieval over official congressional documents (hearings, committee reports, Congressional Record) with citations and govinfo.gov links, designed for grounding AI answers in the official record.

MCP DevTools Server

An MCP server that standardizes and binds development tool patterns, enabling AI assistants like Claude Code to generate code more efficiently with fewer errors and better autocorrection.

GhostDesk

A virtual Linux desktop as an MCP server, shipped in Docker. Agents drive screen, mouse, keyboard across any GUI — browsers, IDEs, office suites, Wine/Windows apps, legacy software — many in parallel.

cork-defi

Enables interaction with the Cork DeFi protocol for reading live chain state, computing bit-exact math, building unsigned bundles and orders, and managing markets, all without signing or broadcasting.