Tavily Search API
TavilyTavily Search API is a hosted search service for AI agents and retrieval-augmented generation workflows, returning structured results through search, extract, crawl, map and research endpoints. Its documentation lists integrations with frameworks such as LangChain and CrewAI.
Openness
1 high confidence- license
- Proprietary(API-only)
- source
- closed(no self-hostable core
- managed
- cloud SaaS with API keys, credits, rate limits
Tavily is sold only as a hosted API, and its terms grant customers no right to the source code behind the service. Only the client libraries are public, under the MIT license, so nothing that runs the search can be self-hosted.
- https://docs.tavily.com/ recorded 2026-09-28
The documentation home is organized around a hosted API ("Get an API key", "Learn how Tavily API credits work.", "Understand Tavily’s rate limits and policies."). Searched the body for self-host, on-prem, open source, GitHub and download: none, so no self-hostable build or source is documented.
- https://raw.githubusercontent.com/tavily-ai/tavily-python/master/LICENSE recorded 2026-09-28
MIT License
- https://www.tavily.com/terms recorded 2026-09-28
any right to obtain possession of any source code, data or other technical material related to the Services
Adoption
4 high confidenceTavily's adoption is measured by the developer count on its own homepage, a figure the vendor reports itself. Nebius's announcement of its agreement to acquire Tavily gives separate user and SDK download figures, which support the same scale but count different things.
- https://nebius.com/newsroom/nebius-announces-agreement-to-acquire-tavily-to-add-agentic-search-to-its-ai-cloud-platform recorded 2026-09-28
more than one million users
- https://www.tavily.com/ recorded 2026-09-28
Trusted by 2M+ developers around the world
Capability
4 medium confidenceTavily's search call can return the page text behind each result directly, through include_raw_content, and include_answer returns a composed answer in the same response. But nothing in that response attributes the answer to the sources beside it, so the answer arrives without citations, which puts it level with Firecrawl.
- https://docs.tavily.com/documentation/api-reference/endpoint/search recorded 2026-09-17
The Tavily Search endpoint reference. The documented request body carries `include_answer` and `include_raw_content` alongside `max_results`, `topic`, `time_range` and the domain filters.
- https://docs.tavily.com/documentation/best-practices/best-practices-search recorded 2026-09-17
Tavily's search best-practices guide: search depths, chunking and `chunks_per_source`, describing what comes back per source. No citation contract is documented for the `answer` field.
Verified 2026-09-17