tavily-mcp

What it does

Tavily MCP provides real-time web search, data extraction, site mapping, and web crawling tools for Claude and other MCP clients. Built on Tavily's search API, it enables AI agents to ground responses in current information, extract structured data from web pages, and systematically explore website architectures. Tools include tavily-search for real-time queries, tavily-extract for intelligent page data extraction, tavily-map for site topology analysis, and a web crawler for deep exploration.

Who it's for

AI application developers and agent engineers integrating real-time web research into Claude workflows. Teams building research assistants, fact-checking systems, and knowledge synthesis tools where current information is essential.

Common use cases

• Fetch real-time information on trending topics to ground AI-generated analysis
• Extract structured data (pricing, product specs, contact info) from competitor or reference websites
• Build knowledge graphs by mapping website hierarchies and link structures
• Systematically crawl documentation or FAQ sites to populate model context windows

Setup pitfalls

• Requires a Tavily API key from tavily.com. Choose between embedding it in the MCP URL or using OAuth authentication; the latter is more secure but requires completing an auth flow in your client.
• Two deployment modes: use the hosted remote MCP at https://mcp.tavily.com/mcp/ (no local setup) or install locally via npx tavily-mcp (requires Node.js). Remote mode is simpler for most use cases.
• The npm package has CI failures—verify the latest release works in your environment before adding to critical workflows.