Brave Search MCP

The Brave Search MCP Server is Brave's official MCP server for its Search API. It gives AI assistants real-time web search plus several specialised searches: local businesses and places, images, videos, news, an AI summariser, and an LLM-context tool that returns pre-extracted web content designed for grounding answers and RAG pipelines.

It suits anyone who wants an assistant to answer with current information rather than training data, especially when you prefer an independent search index over Google or Bing. Web search supports fine-grained parameters such as country, language, result count, freshness filters, safe search and Goggles for custom re-ranking. Version 2 changed defaults: the server now uses stdio by default (set `BRAVE_MCP_TRANSPORT=http` for HTTP), and image search no longer returns base64 image data, which saves a lot of context.

Getting started: sign up for a Brave Search API account, choose a plan and create an API key, then add the server to your client, for example with Docker (`docker run -i --rm -e BRAVE_API_KEY docker.io/mcp/brave-search`) or via npx, passing the key as `BRAVE_API_KEY`.