Tavily
Search APICheapOne API for search/extract/crawl/research — real-time web for agents and RAG.
What it is
Tavily is a real-time search engine for AI agents and RAG: one API covers Search, Extract, Crawl, Map and Research. `/search` — semantic search with ranked, distilled snippets (agents don't waste tokens on noise) and an optional ready `answer`; `/extract` — pulls full content from up to 20 URLs per call (clean text/markdown/images); `/crawl` — smart site traversal respecting robots.txt and dynamic content; `/map` — structure map; the Research API (GA Jan 2026) — deep multi-step research via a single endpoint with multi-agent coordination. A built-in layer blocks PII leakage, prompt injection and malicious sources; SOC 2, zero data retention. In the red* stack it's the core of quick search (redresearch quick mode) and the panels' fact-grounder — its ready answer is the basis of a grounded reply in seconds.
What it can do
- One API: Search + Extract + Crawl + Map + Research
- Ready `answer` + distilled snippets (token-efficient)
- `/extract` up to 20 URLs per call (text/markdown/images)
- `/crawl` — smart traversal with robots.txt
- Research API — deep multi-step research
- Blocks PII/prompt-injection/malicious sources
- SOC 2, zero data retention
- Fast (~1-3s), built for agents/RAG
Pricing
Free tier
1,000 searches/mo free, no card
Paid from
pay-as-you-go, ~$0.008 / поиск
Similar services
| Service | How it differs |
|---|---|
| Tavilypay-as-you-go, ~$0.008 / поиск | One API for search/extract/crawl/research — real-time web for agents and RAG. |
| Exa | Neural search by meaning. |
| Perplexity Sonar | Grounded answer with citations. |
| Serper | Raw Google SERP, cheap. |