AI Potluck
Product / UX / Agent tools & protocols

Tavily Search API

Tavily

Search API purpose-built for AI agents, returning clean, structured results optimized for LLM consumption rather than human browsing. Includes search, extract, crawl, and map endpoints. The default search provider in most agent tutorials and frameworks; integrated into LangChain, CrewAI, and AutoGen out of the box. YC-backed, rapidly growing as the canonical 'give your agent web search' solution.

Proprietary web-access API for agents: search, extract, crawl, map, research endpoints; Python/JS SDKs. PRECISION (verify 2026-06-04): Nebius announced an agreement to acquire Tavily on 10 Feb 2026 for $275M (up to $400M on milestones), expected to close within weeks, an acquisition agreement, not necessarily a fully-closed deal at announcement. Verified live via docs.tavily.com + tavily.com June 2026.

Openness

1 high confidence
1.0
license
Proprietary(API-only)
source
closed(no self-hostable core
managed
cloud SaaS with API keys, credits, rate limits

Closed search/scrape API (the recipe's closed counterpart to Milvus/Qdrant); only client SDKs are open, the search service itself is proprietary.

  • https://docs.tavily.com/ recorded 2026-06-04

    API keys, credits, rate limits, hosted endpoints, no self-hostable open core

Adoption

4 high confidence
4.0

Tavily's own homepage states 2M+ developers and 300M+ monthly requests, billions of pages crawled (both figures re-confirmed on the live homepage 2026-06-04); it is the default search provider wired into LangChain/CrewAI/LlamaIndex agent stacks. Developer reach solidly in 1-10M band; request volume is far higher. Acquisition by Nebius corroborates enterprise traction.

Capability

4 medium confidence
4.0

Full {coverage, freshness, structured output, rate limits} feature set purpose-built for agents and the de-facto default in agent frameworks; near-frontier among agent search APIs but not the single category-definer.

Unchanged since 2026-06-24 (last edited, not re-checked)