Web Search MCP

🚀 Web_Search_MCP

A web search tool server built with Tavily API and MCP (Model Context Protocol)

This project demonstrates how to use the Tavily API to create a web search tool and integrate it with MCP for seamless interaction with AI systems. This enables you to provide real-time web search functionality for your language models or applications.

🚀 Quick Start

Prerequisites

Before running the project, ensure you have the following:

• Python 3.8+: Python 3.8 or a higher version installed on your system.
• Tavily API Key: A valid Tavily API key, which can be obtained by registering on the Tavily website.
• uv: The uv package manager for Python. You can install it using the following command:

pip install uv

Installation

1. Create the project directory: Create a project directory and navigate to it:

uv init Web_Search_MCP    

*** By running the above code, uv will create the pyproject.toml and .venv files *** 2. Create project files: Create the files main.py and .env in the Web_Search_MCP directory. 3. Activate the virtual environment: Navigate to .venv/Scripts/ and activate the virtual environment. 4. Copy the code: Copy the code into the newly created main.py and .env files. 5. Install dependencies: Use uv to install the required Python packages:

uv add "mcp[cli]" python-dotenv langchain-community tavily-python

Configuration

1. .env file:
   • Create a file named .env in the Web_Search_MCP directory.
   • Add your Tavily API key to the .env file in the following format:

TAVILY_API_KEY='your_tavily_api_key'

Replace your_tavily_api_key with your actual Tavily API key.

2. claude_desktop_config.json:
   • Create a file named claude_desktop_config.json in the Web_Search_MCP directory.
   • Paste the following content into the file:

{
    "proxy": null,
    "model_name": "Claude 2",
    "temperature": 0.1,
    "max_tokens": 1000,
    "tools": [
        {
            "type": "function_calling",
            "tool_name": "search_web",
            "description": "Use this tool to search the web and obtain relevant information."
        }
    ]
}

• Save the file.

Running the Project

1. Install dependencies: Open a terminal in the project directory and run:

pip install -r requirements.txt

2. Start the server: Run the following command to start the server:

uvicorn main:app --reload

3. Access the interface: Open a browser and visit http://localhost:8000.

✨ Features

The core of this project is the search_web tool, which offers the following features:

• Web search: Accepts a search query as input and uses the Tavily API to retrieve relevant search results from the web.
• Detailed results: Provides detailed information from the search results, including website content, URLs, relevance scores, content types, and direct answers (if available).
• Formatted output: Returns search results in a well-structured JSON format. The output includes a status indicator (success or error), an array of results on success, and a timestamp.
• Error handling: Gracefully handles errors during the search process and returns detailed error information in JSON format.
• Asynchronous processing: The search tool is asynchronous and can handle multiple requests simultaneously.

📚 Documentation

Project Overview

The Web_Search_MCP project utilizes the following key components:

• Tavily API: A powerful search API that provides real-time, comprehensive web search results, including answers, raw content, and relevant metadata.
• FastMCP: The FastMCP class uses Python type hints and docstrings to automatically generate tool definitions, making it easy to create and maintain MCP tools.
• Langchain: Specifically, the TavilySearchResults tool provided by Langchain is used to interact with the Tavily API efficiently.
• Dotenv: A library for loading environment variables from a .env file, securely managing sensitive information such as API keys.
• uv: A very fast Python package installer and resolver used to manage and run this project.

Tool Usage

Once the server is running, it exposes a tool named search_web according to the model context protocol.

• Tool name: search_web
• Input:
  • query (str): The search query to be submitted to the web.
• Output: A structured JSON string containing the search results.

Example JSON response (success):

{
    "status": "success",
    "results": [
        {
            "title": "...",
            "url": "...",
            "content": "text",
            "score": 0.9,
            "raw_content": "..."
        },
        {
            "title": "...",
            "url": "...",
            "content": "text",
            "score": 0.9,
            "raw_content": "..."
        }
    ],
    "timestamp": 1708849844.064655
}

📄 License

No license information provided in the original document, so this section is skipped.

🔧 Technical Details

No specific technical implementation details (more than 50 words) provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

# There is no specific code example for basic usage in the original document.
# If there were, the code would be placed here as it is.

Advanced Usage

# There is no specific code example for advanced usage in the original document.
# If there were, the code would be placed here as it is.