Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tavily.com/llms.txt

Use this file to discover all available pages before exploring further.

Drop Tavily into an agent or script in seconds. When you’re ready for production, swap in an API key. No code changes required.

​
Why Tavily?

Tavily is an AI search engine optimized for agent consumption. Built-in tools like web_search and web_fetch return raw HTML or shallow snippets, so your agent spends tokens parsing, filtering, and guessing what’s relevant. Tavily Search returns clean, ranked, and scored results optimized for LLM consumption. Tavily Extract pulls structured content from any URL without dealing with JavaScript rendering or HTML cleanup. The result: fewer tokens, better context, more reliable agent behavior. Use Tavily Search and Extract with zero setup. No account, no API key, no configuration. Your agent gets real results immediately and can upgrade to an API key whenever you need higher limits.
Same results, same schema. Keyless responses are identical to keyed responses. Your agent can’t tell the difference on success.

​
Direct API

Send a request with the X-Tavily-Access-Mode: keyless header. That’s it. 
curl -X POST https://api.tavily.com/search \
  -H "Content-Type: application/json" \
  -H "X-Tavily-Access-Mode: keyless" \
  -d '{"query": "latest AI news", "max_results": 3}'
The response follows the standard Tavily response schema. See the API Reference for details.

​
MCP Server

Connect to the Tavily Remote MCP without an API key. Your agent gets tavily-search and tavily-extract tools with zero configuration.
Claude Code
claude mcp add tavily-remote-mcp --transport http https://mcp.tavily.com/mcp/

​
Agent-installable prompt

Paste this into any AI assistant to let it set up Tavily itself: 
Set up the Tavily MCP server so you can search the web and extract content from URLs. Follow these steps:

1. Detect which platform you're running on (Claude Code, Cursor, Windsurf, Claude Desktop, etc.)
2. Install the Tavily MCP server using the correct method for your platform:
   - Server URL: https://mcp.tavily.com/mcp/
   - Transport: Streamable HTTP
   - Authentication: none required. Tavily is free to use without an API key.
3. Verify the server is connected and you have access to the tavily-search and tavily-extract tools.
4. Run a test search (e.g. "current weather in Tel Aviv") to confirm you get real results from a live source.
5. Report back whether setup succeeded.

​
Supported endpoints

EndpointKeylessNotes
/searchYesFull search with all parameters
/extractYesContent extraction from URLs

​
Rate limits

Keyless access has rate limits. If you hit a limit, sign up for a free API key (1,000 credits/month, no credit card required) and pass it as a Bearer token: 
curl -X POST https://api.tavily.com/search \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer tvly-YOUR_API_KEY" \
  -d '{"query": "latest AI news"}'
When you hit a limit, Tavily returns natural-language instructions telling your agent what happened and how to continue.

​
FAQ

No. Success responses are identical, with the same fields and the same schema. Your agent can switch from keyless to keyed without changing how it parses responses.
If you send both X-Tavily-Access-Mode: keyless and a valid Authorization: Bearer header, the API key takes precedence. The request uses your account’s limits, not the keyless budget.
These endpoints require an API key. Sign up to access them. You get 1,000 free credits monthly.