Brave Search API

Executes real-time queries against a 30+ billion page index and aggregates results from multiple sources, returning up to 5 snippets per result with metadata (URLs, titles, publication dates). Uses a distributed crawling and indexing architecture to maintain freshness without relying on cached or stale indices. Results are optimized for RAG pipelines by structuring snippets for LLM consumption.

Processes natural language queries through an LLM-powered summarization endpoint that generates concise, factual answers grounded in multiple web sources. Citations are automatically generated and linked to source documents, reducing hallucination by constraining the model to verifiable facts. Streaming is supported for real-time response delivery, and output is optimized for conversational interfaces.

Executes searches without building user profiles, tracking search history, or using behavioral data for ranking or personalization. The implementation avoids storing personally identifiable information, using cookies for tracking, or selling user data to third parties. Privacy is enforced at the infrastructure level through data minimization and anonymization.

Provides a free tier with $5 in monthly auto-credited API usage, allowing developers to experiment with Brave Search without upfront payment. The credit resets monthly and covers both Search and Answers endpoints at their respective per-request rates. Exact request quotas for the free tier are not documented, but the $5 credit translates to approximately 1,000 Search requests or 1,250 Answers requests per month.

Provides a drop-in compatible interface with OpenAI's chat completion API, allowing developers to swap Brave Answers for GPT-4 in existing codebases with minimal changes. The endpoint accepts OpenAI-formatted requests (messages array, model parameter) and returns OpenAI-compatible response objects, enabling seamless integration into LLM applications already using OpenAI SDKs.

Brave Search is natively integrated as a tool within Claude's Model Context Protocol, allowing Claude to invoke Brave Search directly from conversations without requiring manual API integration. The integration exposes Search and Answers endpoints as callable tools with automatic parameter binding, enabling Claude to autonomously decide when to search the web for information.

Executes location-aware searches that return results filtered by geographic proximity, enabling queries for local businesses, services, and events. The implementation uses geolocation data (IP-based or explicit coordinates) to rank and filter results by distance, returning location metadata (addresses, phone numbers, hours) alongside web results.

Executes image and video searches against a visual index, returning results with thumbnails, source URLs, and metadata. The implementation indexes images and videos from web crawls, enabling searches for visual content without relying on third-party image APIs. Results include image dimensions, alt text, and source page context.

Executes news-specific searches that return articles from news sources with publication dates, author information, and content snippets. The implementation maintains a separate news index with temporal filtering, enabling queries for recent news, historical articles, or news within a specific date range. Results are ranked by recency and source authority.

Provides a domain-level filtering and reranking system called 'Goggles' that allows users and developers to customize search results by blocking/prioritizing specific domains or content types. The implementation operates as a post-processing layer on search results, enabling custom ranking algorithms without modifying the underlying index. Goggles can be shared and versioned, supporting collaborative filtering.

Supports HTTP streaming (Server-Sent Events or chunked transfer encoding) on the Answers endpoint, enabling real-time delivery of summarized responses as they are generated. The implementation streams answer text incrementally, allowing clients to display partial results before the full response is complete. Streaming reduces perceived latency and enables interactive conversational interfaces.

Provides enterprise-tier support with custom API endpoints, dedicated infrastructure, and negotiated volume pricing. Enterprise customers can request custom endpoint configurations, higher rate limits, and dedicated support. Pricing is custom-negotiated based on volume and use case, with no public rate card.