Summary

Search and the reading that follows it, from one place.

What it takes off you is the fetch-and-parse step after a search. The URLs a search returns can be extracted as markdown, or the whole section of the site around them crawled under bounds you set, without a second tool and a second set of failures to handle.

What it is

A search server whose five calls escalate. Search returns snippets and URLs, extract pulls a page's content, crawl walks outward from a starting page, map returns the URL list alone, and research answers a question across sources.

What you get

Web search with topic, depth, date range, included and excluded domains, country, and optional images
Page content extracted from a list of URLs as markdown or text, at the depth you choose
A crawl from a starting URL, bounded by breadth, depth and path or domain selection, and steered by an instruction
A site's structure mapped as a list of URLs under those same bounds
A research call that gathers across sources and returns a written answer

Requirements

Npx on your PATH, and a Tavily API key in TAVILY_API_KEY.

Setup effort

One command plus a key — claude mcp add --transport http tavily https://mcp.tavily.com/mcp, then supply credentials