Tavily
A search index native to agents and RAG, returning snippets with relevance scores and an optional answer.
Where Tavily lands
Composite scores out of 10, from benchmark runs where every provider faces the identical task corpus. Rank is within that capability's field; the newest run backing this page is 2026-06-22. Full method: how we measure.
- Coverage
- 0.78
- Latency p50
- 174 ms
- Cost
- $0.014 / useful
- Errors
- 0%
- Correctness
- 1.00
- Citation faithfulness
- 0.24
- Latency p50
- 1.4 s
- Cost
- $0.008 / call
- Errors
- 0%
- Recall@10
- 0.58
- Latency p50
- 2.1 s
- Cost
- $0.014 / useful result
- Errors
- 0%
The honest pitch
Pick it when you're building the typical agent or RAG stack and want relevance scores plus an optional answer from a single call.
Not the pick if per-result dates matter, or if a documented SLA/compliance posture is required.
About Tavily
Tavily approaches search from the agent side: the API was designed from scratch for retrieval-augmented generation and AI agents. Each call returns clean results carrying relevance scores, and can attach an LLM-ready answer plus full page content in the same response — the exact shape a RAG pipeline consumes, leaving no post-processing to do. Built squarely for grounding and trusted widely across the agent world, it has settled in as a default search layer across source-backed AI applications.
Reaching it through NativePort
Identical paths, parameters and responses to Tavily's own documentation — the gateway holds the upstream credential and meters your balance at Tavily's real published usage price.
curl https://api.nativeport.ai/tavily/<native-path> \ -H "Authorization: Bearer $NATIVEPORT_API_KEY"