The server exposes several specialized search tools. brave_web_search performs comprehensive web searches with rich result types and advanced filtering including country, language, safe search, freshness, and custom Goggles re-ranking definitions. brave_local_search finds local businesses and places with detailed ratings, hours, and AI-generated descriptions. brave_image_search and brave_video_search provide targeted media search with comprehensive metadata. brave_news_search surfaces current events, while brave_summarize delivers AI-generated summaries grounded on search results for reduced hallucination.

Version 2.x improved performance by removing base64-encoded image data from responses, returning a leaner response object that closely mirrors the original Brave Search API structure. The server supports both STDIO (default) and HTTP transports, configurable via the BRAVE_MCP_TRANSPORT environment variable or --transport runtime argument. Installation is available through npm (npx @anthropic-ai/brave-search-mcp) or Docker, and it requires a Brave Search API key from the developer dashboard.

This MCP server is particularly valuable for AI agents that need grounded, real-time information retrieval. By providing search results through the standardized MCP protocol, it enables any MCP-compatible client — Claude Desktop, Cursor, VS Code Copilot, Windsurf, and others — to perform web research without custom API integration code. The Pro plan unlocks additional features like extra snippets and full local search capabilities.