Brave Search API
A search API that returns results from Brave’s own web index, with no fallback to Google or Bing, and an LLM Context endpoint made for grounding agents.
Plans
| Plan | What it gives |
|---|---|
| Search | Web results plus LLM-optimized context, Goggles for custom reranking and filtering, extra alternate snippets, schema-enriched results; highest rate limit of the self-serve plans |
| Answers | Summarized answers grounded in one or more searches, with citations, streaming and OpenAI SDK compatibility; billed per request plus tokens; lower rate limit |
| Enterprise | Custom terms, capacity and endpoints, full-funnel zero data retention, custom agreements and invoicing; through sales |
Billing is prepaid credits with no subscription, and both self-serve plans include a monthly free credit. The page quotes two different free-credit figures (a headline “up to 5 on the plan sections), so check the plans page for the current amounts.
Endpoints
LLM Context (agent grounding), Answers, Web search, News, Images, Videos and Places/POIs; spellcheck is a request parameter.
Points that matter for agents
- Own index: the page says results come from Brave’s independent index, not a wrapper around another engine
- Zero data retention: described as architectural, with queries staying on infrastructure Brave operates; full-funnel ZDR is part of Enterprise
- Storage and training: storing results or using them to train models needs a custom plan with storage rights
- MCP: a Brave Search MCP server is published on GitHub
- Goggles: user-defined ranking rules to boost or drop domains, which is rare among search APIs
Related
- ai-search-apis-overview
- brave-leo (Brave’s browser assistant, a different product)
- serper-dot-dev, exa-search-api, parallel-search-api, tavily-search-api