Brave Search MCP

Brave Search MCP is a Model Context Protocol server that connects an AI assistant to the Brave Search API. It gives a model live web results from an independent index, which matters for two reasons: the answers are not limited to what the model absorbed during training, and the underlying index is built by Brave rather than syndicated from another engine. The server exposes two kinds of lookup. Web search takes a query and returns ranked results with titles, URLs, and snippets, with paging and result count controls so an agent can go deeper on a topic instead of stopping at the first page of hits. Local search resolves place-oriented queries, returning business listings with address, category, rating, and opening hours where the index carries them, and falling back to standard web results when a query turns out to have no local intent. Both run through a Brave Search API key, which is what meters usage and separates this from scraping a search results page. This server began as one of the reference implementations published alongside the Model Context Protocol specification. That reference copy sits in the archived servers repository, with maintenance continuing in Brave's own distribution. It is a common first integration for anyone building a research or fact-checking agent, and a useful reference for developers who want to read a small, complete server implementation before writing their own.