Tavily MCP server
Last verified: 2026-09
Search API built for LLM retrieval.
Community Browser & search Needs a secret Claude Desktop, Cursor, Windsurf, Claude Code
What it does
Tavily's search API for agents. Good default when you want cited web results without driving a browser. Remote MCP is on their docs; the npx package is the local path.
It sits in Browser & search: Three different jobs get mixed up here.
Good for
- Cited web results for a research question in Claude Desktop.
- Ground a claim with sources without driving a browser.
- One retrieval API for an agent that must not invent URLs.
- Extract page text when the connected Tavily server exposes it.
Tools
- search — Web search with snippets and URLs.
- extract / crawl (package-dependent) — Pull page text when the connected server exposes it — check the current tool list after connect.
Config (Claude Desktop / Cursor / Windsurf)
Paste the JSON below. Same mcpServers shape. Replace placeholder paths and secrets.
Windows paths look like C:\\Users\\you\\project, not /path/to.
Host-side steps →
Claude Desktop
| OS | Config file |
|---|---|
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Windows | %APPDATA%\Claude\claude_desktop_config.json
(usually C:\Users\<you>\AppData\Roaming\Claude\) |
| Linux | ~/.config/Claude/claude_desktop_config.json |
Cursor
| Scope | macOS / Linux | Windows |
|---|---|---|
| This project | .cursor/mcp.json in the repo root | |
| This user | ~/.cursor/mcp.json |
%USERPROFILE%\.cursor\mcp.json |
Windsurf
| OS | Config file |
|---|---|
| macOS / Linux | ~/.codeium/windsurf/mcp_config.json |
| Windows | %USERPROFILE%\.codeium\windsurf\mcp_config.json |
{
"mcpServers": {
"tavily": {
"command": "npx",
"args": [
"-y",
"tavily-mcp"
],
"env": {
"TAVILY_API_KEY": "tvly-..."
}
}
}
}Windsurf remote MCP: use serverUrl instead of url if the block below is HTTP.
One-liner: npx -y tavily-mcp
Secrets it wants: TAVILY_API_KEY
How to get started
- Create a Tavily key. Set TAVILY_API_KEY.
- Restart and ask for a search with sources.
- Prefer this over Playwright for 'what does the web say' questions.
Access risk
The key is spend + whatever query text you send to Tavily.
When to skip it
Skip it for a single known URL (Fetch) or when you need a logged-in page (Playwright).
Instead: Exa, Brave Search, Fetch.
Vs alternatives
| Server | Official? | Needs a secret? | Best for |
|---|---|---|---|
| Tavily | No | Yes | Search API built for LLM retrieval. |
| Exa | No | Yes | Neural web search aimed at agents. |
| Brave Search | Yes | Yes | Web search via the Brave Search API. |
| Fetch | Yes | Yes | Fetch a URL and turn the page into readable text. |
More in Browser & search
FAQ
Tavily or Exa?
Same job, different vendor. Install one. Tavily markets itself as LLM retrieval; Exa as neural search.
Remote or npx?
Tavily documents a remote MCP. The npx package is the local path with TAVILY_API_KEY.
Good for a logged-in page?
No. Use Playwright. Tavily is public web.
What leaks to Tavily?
The query text and your key. Do not search customer PII.