Skip to main content
Guide1 min read

Tavily MCP server: web search & extract for AI agents (2026)

Add real-time search, extract, map and crawl to Claude or Cursor with the official Tavily MCP server. API key, npx or remote setup, and fixes.

Tavily is a search API built for AI agents, and its official MCP server exposes search, content extraction, site mapping and crawling — clean, structured web data your agent can act on. Here's the setup.

Step 1 — get an API key

Sign up at Tavily and copy your API key (starts with tvly-). There's a free monthly credit tier.

Step 2 — configure the client

Local via npx:

{
  "mcpServers": {
    "tavily": {
      "command": "npx",
      "args": ["-y", "tavily-mcp@latest"],
      "env": {
        "TAVILY_API_KEY": "tvly-your_key"
      }
    }
  }
}

Tavily also hosts a remote endpoint you can add as an OAuth/URL connector if your client supports remote MCP. Restart the client; see MCP config file location.

What it exposes

  • search — real-time web results tuned for LLMs.
  • extract — clean content from specific URLs.
  • map / crawl — discover and pull a site's pages.

Verify

Ask: "Use Tavily to find and extract the 3 best guides on MCP security, with sources." Structured results back confirm it works.

Common problems

  • 401 / invalid key — recheck the tvly- key.
  • Quota exceeded — you used the free credits; add billing or wait for reset.
  • No tools in client — Node 18+ for npx; see Cursor MCP not working.

Going further

Tavily pairs with Sequential Thinking and a filesystem server for research agents. Compare with Perplexity (synthesized answers) and Bright Data (heavy scraping/unblocking). Browse the search category or curated loadouts.

Loadout

Build your AI agent loadout

The directory of MCP servers and AI agents that actually work. Pick the right loadout for Slack, Postgres, GitHub, Figma and 20+ integrations — with install commands ready to paste into Claude Desktop, Cursor or your own stack.

© 2026 Loadout. Built on Angular 21 SSR.