Tavily MCP server
Official Tavily server giving agents real-time web search, page extraction, site crawling, and site mapping built for AI.
Tavily MCP is the official server from Tavily, a search and web-access API designed for LLMs rather than humans. It exposes four focused tools: a real-time web search optimized for agent consumption, an extract tool that pulls clean content from specific pages, a crawl tool that systematically explores a site, and a map tool that produces a structured map of a website's pages. Together they cover the common loop of an agent that needs to find sources, read them, and traverse a domain without bolting on a separate scraping stack.
It runs two ways. Locally it installs over stdio with npx using the tavily-mcp package and a TAVILY_API_KEY env var. Tavily also hosts a remote Streamable HTTP endpoint at https://mcp.tavily.com/mcp/ where the key is supplied as a tavilyApiKey query parameter; the remote endpoint additionally supports OAuth for clients that handle it automatically, so the key need not be embedded in the URL. The API key authenticates requests and governs usage limits.
Quick install
Copy-paste configs are provided for all 8 supported clients. Pick your client below.
Add to ~/.claude.json
{
"mcpServers": {
"tavily": {
"type": "http",
"url": "https://mcp.tavily.com/mcp/"
}
}
}claude mcp add --transport http tavily https://mcp.tavily.com/mcp/Heads up
- First tool call opens a browser to authorize.
Available tools
| Tool | Description |
|---|---|
| tavily-search | Real-time web search that returns results optimized for LLM consumption. |
| tavily-extract | Extract clean, structured content from one or more specific web pages. |
| tavily-crawl | Systematically crawl a website, following links to gather pages across the domain. |
| tavily-map | Generate a structured map of a website's pages and their relationships. |
Required configuration
- TAVILY_API_KEYRequired
Tavily API key used to authenticate requests and govern usage limits.
What you can do with it
Real-time answers with sources
An agent runs tavily-search to get current, citation-ready results tuned for LLMs, then reasons over them without parsing a raw SERP.
Read and traverse a target site
Pair tavily-extract to pull clean content from known pages with tavily-crawl and tavily-map to discover and walk the rest of a domain.
FAQ
- Is it free?
- The server is open source and free to run; calls to Tavily's API require a TAVILY_API_KEY and consume credits. Tavily offers a free tier with a monthly credit allowance for getting started.
- Does it support remote/OAuth?
- Yes. Tavily hosts a remote Streamable HTTP endpoint at https://mcp.tavily.com/mcp/. You can pass the key as a tavilyApiKey query parameter, or use OAuth, which compatible clients handle automatically so the key is not embedded in the URL. The local stdio package uses the key via an env var.