brave-search

This MCP server provides Brave Search powered capabilities via a single MCP endpoint. It exposes tools for web search, image search, video search, news search, local points of interest search, and an LLM-context search that fetches and optimizes page content for AI pipelines. You can call the built-in tools such as brave_web_search, brave_image_search, brave_news_search, brave_video_search, brave_local_search, brave_llm_context_search to perform targeted queries. By running the server in HTTP/streamable mode, you can access the MCP at the configured host and port and optionally expose a UI for tool configuration. The server expects a Brave API key (BRAVE_API_KEY) when operating in HTTP mode, and supports standard options like port and host as well as a allows-list for host validation.

Prerequisites: Node.js (recommended LTS, v18+), npm or corepack. You do not need to install the package globally if you use npx. Steps:\n1) Install Node.js from https://nodejs.org if not already installed.\n2) (Optional) Verify npm is available: npm -v.\n3) Run the MCP server in HTTP/streamable mode by supplying your Brave API key:\n\nBRAVE_API_KEY="your_key_here" npx -y brave-search-mcp --http\n\nTo run with a UI (if supported):\nBRAVE_API_KEY="your_key_here" npx -y brave-search-mcp --http --ui\n\nEnvironment variables you may configure:\n- BRAVE_API_KEY: Brave Search API key (required for HTTP mode).\n- PORT: HTTP port (default 3001).\n- HOST: Interface to bind to (default 0.0.0.0).\n- ALLOWED_HOSTS: Comma-separated allowed hostnames for Host header validation.\nYou can also use a non-HTTP (stdio) mode by omitting the --http flag.

Tips and common issues:\n- Ensure BRAVE_API_KEY is set when using HTTP mode; the server will not fetch results without it.\n- If you’re behind a firewall or reverse proxy, configure HOST and ALLOWED_HOSTS to match your network.\n- The brave_local_search tool falls back to brave_web_search if no local results are found; consider providing a relevant query to maximize hits.\n- For large responses, consider adjusting maximum tokens/snippets parameters in your downstream application if using the LLM context search.\n- If you encounter rate limits, review your Brave API plan and consider staggering requests or caching results on your side.