Tavily MCP Server

This repository hosts a Model Context Protocol (MCP) server that leverages the Tavily search API to provide AI-powered web search capabilities for Large Language Models (LLMs). It enables LLMs to perform web searches, obtain direct answers, and search recent news with AI-extracted content.

Features

The server offers three primary tools: tavily_web_search for general web searches, tavily_answer_search for direct answers with supporting evidence, and tavily_news_search for recent news articles. Each tool accepts a search query and optional parameters like max_results, search_depth, and domain filters. Prompt templates are also provided for each search type.

Prerequisites & Installation

The server requires Python 3.11+, a Tavily API key, and uv (recommended). Installation can be done via pip or uv, or directly from the source code. The README provides detailed instructions for each method, including setting up a virtual environment and installing dependencies.

Configuration & Usage

Configuration involves setting up the Tavily API key through a .env file, environment variable, or command-line argument. The README also provides configuration instructions for Claude.app. Usage examples demonstrate how to perform web searches, generate reports with domain filtering, obtain direct answers, and search for news.

Testing & Contributing

The project includes a test suite that can be run using uv sync --dev and ./tests/run_tests.sh. Contributions are welcome via forking, creating a feature branch, and submitting a pull request. The project is licensed under the MIT License.