tavily-mcp

What it does

The Tavily MCP server integrates real-time web search and data extraction into Claude and other MCP clients. It provides four core capabilities: a search tool for finding current information across the web, an extract tool for pulling specific data from web pages, a map tool for creating structured representations of website layouts, and a crawl tool for systematically exploring multi-page sites. The server connects to Tavily's hosted API via their remote MCP endpoint or an optional npm-based local installation. Authentication is handled through API keys or OAuth, with support for setting default parameters like search depth and result limits.

Who it's for

Developers building research assistants in Claude Code or Cursor, knowledge workers who need real-time web context for competitive analysis and reporting, and engineers creating AI applications that require live data extraction, web scraping, or structured information gathering from websites.

Common use cases

• Search the web from Claude for fact-checking and retrieving the latest information
• Extract structured data from web pages including pricing, reviews, product details, and contact information
• Map website information architecture and structure before systematic crawling operations
• Gather data across multiple pages for competitive analysis, market research, or knowledge graph construction

Setup pitfalls

• Requires a Tavily API key obtained from tavily.com before configuring the server
• Uses network calls to Tavily's remote servers—no option for fully local operation
• OAuth authentication may require additional configuration steps in certain MCP clients