Overview
The Brave Search API is a developer product built on Brave’s own crawler and index of more than 30 billion pages, refreshed by over 100 million page updates a day. Its pitch is independence: results do not depend on Google’s or Bing’s ranking, and because Brave operates the entire search stack it can offer true Zero Data Retention across endpoints. The API surface goes well beyond web search and includes News, Images, Videos, Local/POI, Autosuggest, Spellcheck, a summarizer, an LLM Context endpoint tuned for RAG pipelines, and Answers, a grounded question-answering product with citations and streaming. Pricing is arranged into self-serve Search, Answers, Spellcheck, and Autocomplete plans, each carrying $5 in free monthly credits (roughly $20 across all four). In a June 2026 head-to-head evaluation Brave reported its Ask Brave assistant, running on the LLM Context grounding endpoint with an open-weights Qwen3 model, outscored ChatGPT, Perplexity, and Gemini on a 1,500-query benchmark — a result it credits to higher-quality grounding data rather than a stronger model.
Key Features
- Independent web index of 30B+ pages — results don’t depend on Google’s or Bing’s ranking, with Zero Data Retention (no queries stored)
- Structured JSON over REST for Web, News, Images, Videos, Local/POI, plus Autosuggest and Spellcheck endpoints
- LLM Context endpoint purpose-built for RAG — returns pre-extracted, provenance-tagged chunks, tables, and code instead of a raw link list
- Answers plan delivers grounded, cited responses (94.1% SimpleQA F1) at $4/1k searches plus $5/M tokens
- Self-serve API key, $5 monthly free credit per plan (up to ~$20 total), 50 queries/sec on Search — real production capacity
- SOC 2 Type II, Skills support, and an API Assistant in the developer portal
Pricing
| Plan | Price | For |
|---|---|---|
| Free credits | $5 / mo per plan | Prototyping (up to ~$20/mo across all plans) |
| Search | $5 / 1,000 requests | Web/News/Images/Videos JSON + LLM Context |
| Answers | $4 / 1,000 + $5 / M tokens | Grounded, cited responses |
| Autocomplete / Spellcheck | $5 / 10,000 requests | Suggestions and spellcheck as a service |
Plans and per-thousand rates change; confirm on brave.com/search/api before production use.
Comparison
vs. Perplexity: Perplexity is a consumer answer engine you query directly; the Brave Search API is infrastructure you embed. If you need raw, citable web results as JSON to ground your own model, Brave is the building block, while Perplexity is the finished product.
vs. Tavily: Tavily is a search API purpose-built for LLM context, similar in spirit to Brave’s LLM Context endpoint. Brave’s edge is its own full-scale index and higher self-serve throughput (50 q/s), whereas Tavily leans fully into agent-friendly response shaping.
vs. Exa: Exa indexes the web with embeddings for neural/semantic search. Brave returns traditional ranked results from a keyword index plus an LLM Context layer — pick Brave for breadth and ZDR compliance, Exa for embedding-driven relevance.