Tavily MCP server
Real-time web search, extract, map, and crawl for agents.
Tavily's MCP server gives an agent a search engine built for LLMs: real-time web search, clean content extraction, site mapping, and systematic crawling. It's the retrieval layer for research and grounding agents that need fresh, citable web data.
Who it’s for: Research and RAG agents that need current, structured web information — not a stale training cutoff.
Local MCP server — runs on your machine via the command below.
What is Tavily MCP?
Tavily's MCP server gives an agent a search engine built for LLMs: real-time web search, clean content extraction, site mapping, and systematic crawling. It's the retrieval layer for research and grounding agents that need fresh, citable web data.
Who is Tavily for?
Research and RAG agents that need current, structured web information — not a stale training cutoff.
How do I connect Tavily to my agent?
Tavily runs locally over stdio. Add it to your MCP client config: claude mcp add tavily -e TAVILY_API_KEY=tvly-xxx -- npx -y tavily-mcp@latest.
Does Tavily require authentication?
Yes — Tavily authenticates with API key. You connect your own account, and it acts only within the scopes you grant.
How much does Tavily cost?
Free tier · usage-based.
Is Tavily verified on MCPExplorer?
Yes — Tavily is hand-checked and featured on MCPExplorer. We pull its real tool surface, required scopes, and risk labels from the live index so you can see exactly what it does before connecting.
Other MCP servers worth a look.
Tavily is featured on MCPExplorer. Add this badge to Tavily’s site or README — it links back to this page.
HTML — for your website
Markdown — for a README
Set up Tavily and put it to work — Free tier · usage-based. Or see its full verified profile.