
Brave Search
OfficialOfficial Brave Search MCP server: web, local, image, video, news search and AI summarization.
Add to your client
Copy the config for your MCP client and paste it into its config file.
npx -y @brave/brave-search-mcp-serverPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"brave-search": {
"command": "npx",
"args": [
"-y",
"@brave/brave-search-mcp-server",
"--transport",
"stdio"
],
"env": {
"BRAVE_API_KEY": "<your-brave-api-key>"
}
}
}
}Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- Node.js 22.x or later (or Docker)
- A Brave Search API key from https://brave.com/search/api/ (set as BRAVE_API_KEY)
- An MCP-compatible client such as Claude Desktop, VS Code, or MCP Inspector
About Brave Search
The Brave Search MCP Server is Brave's official Model Context Protocol integration for the Brave Search API. It gives AI assistants a privacy-respecting, independently indexed search backend covering web, local business, place, image, video, and news results, plus AI-powered summarization.
Beyond plain search, it exposes a brave_llm_context tool that returns pre-extracted page content tuned for RAG, and a brave_summarizer tool that synthesizes an answer from search results. Local and place searches surface points of interest and business details for location-aware queries.
The server defaults to stdio transport for use with desktop MCP clients, but version 2.x can also run as a streamable HTTP server via --transport http (configurable host and port) for remote or shared deployments. Docker and npx run methods are both documented.
Tools & capabilities (8)
brave_web_searchGeneral web search with filtering options.
brave_local_searchFind local businesses and places.
brave_image_searchSearch for images with metadata.
brave_video_searchSearch for video content.
brave_news_searchSearch current news articles.
brave_place_searchFind points of interest in a geographic area.
brave_llm_contextReturn pre-extracted web content optimized for AI/RAG.
brave_summarizerGenerate an AI summary from search results.
When to use it
- Use it when you want an assistant to search the live web with an independent, privacy-focused index instead of a major engine.
- Use it when you need location-aware answers about local businesses, places, or points of interest.
- Use it when you want image, video, or news results surfaced directly into a chat.
- Use it when you want RAG-ready page content (brave_llm_context) without scraping pages yourself.
- Use it when you want a one-call AI summary synthesized from fresh search results.
Quick setup
- 1Get a Brave Search API key at https://brave.com/search/api/.
- 2Add the server to your client config using npx (`npx @brave/brave-search-mcp-server`) or Docker.
- 3Set the BRAVE_API_KEY environment variable.
- 4Optionally pass `--transport http` with a host/port to run as a remote HTTP server.
- 5Restart your MCP client and verify with a test web search.
Security notes
Requires a Brave Search API key passed via BRAVE_API_KEY; treat it as a secret and avoid committing it to source control. Search results are arbitrary third-party web content, so downstream tools should treat returned text as untrusted.
Brave Search FAQ
Is there a free tier?
Brave offers a free Search API plan with a monthly query allowance and rate limits, plus paid tiers for higher volume. Check brave.com/search/api for current quotas, as they vary by plan.
Can it run as a remote HTTP server?
Yes. By default it uses stdio, but version 2.x supports `--transport http` with a configurable host (default 0.0.0.0) and port (default 8000).
Which clients does it support?
The README documents Claude Desktop, VS Code, and MCP Inspector, with both npx and Docker run options. Any standard MCP client works.
What Node version do I need?
Node.js 22.x or higher is required for the local npx setup; alternatively run it via the published Docker image.
Alternatives to Brave Search
Compare all alternatives →Official MCP reference server that fetches a URL and returns its content as clean Markdown, with chunking.
Official Firecrawl MCP server — scrape, crawl, map, search, and structured extraction for any LLM client.
Exa's neural web search and crawling MCP server — runs locally via npx or as a hosted remote endpoint.
Compare Brave Search with: