The Tavily MCP Server is a Model Context Protocol server that exposes three Tavily-powered search tools to LLMs: tavily_web_search, tavily_answer_search, and tavily_news_search. It enables AI-assisted web searching, direct answers with evidence, and recent-news lookups with publication dates. Each tool accepts a query and optional controls such as max_results (default 5, max 20), and search_depth (basic or advanced). The web search and answer search also allow include_domains and exclude_domains filters; the news search adds days to limit the lookback (default 3). The server provides prompt templates for each search type to guide how the LLM should query Tavily and present results. Prerequisites include Python 3.11+, a Tavily API key, and uv. Installation can be via pip or uv, or from source, and the repository includes Docker deployment options and integration prompts for VS Code. A deprecation note indicates that Tavily’s official MCP server supersedes this project, but the code remains a reference implementation with domain filtering, AI content extraction, and structured responses suitable for MCP workflows. It also supports environment-based configuration via TAVILY_API_KEY and command-line access.