Brave Search MCP server

Connect Brave Search

{"mcpServers": {"brave-search": {"command": "docker", "args": ["run", "-i", "--rm", "-e", "BRAVE_API_KEY", "docker.io/mcp/brave-search"], "env": {"BRAVE_API_KEY": "YOUR_API_KEY_HERE"}}}}

{"servers": {"brave-search-mcp-server": {"command": "npx", "args": ["-y", "@brave/brave-search-mcp-server", "--transport", "stdio"], "env": {"BRAVE_API_KEY": "YOUR_API_KEY_HERE"}}}}

Create the API key in the Brave Search API dashboard. To limit what the agent can call, set BRAVE_MCP_ENABLED_TOOLS or BRAVE_MCP_DISABLED_TOOLS to a space-separated list of tool names.

Or paste this into your coding agent: Connect the Brave Search MCP server to my coding agent using its documented setup: https://github.com/brave/brave-search-mcp-server. Copy

What the Brave Search MCP server does

The Brave Search MCP server connects an AI agent to the Brave Search API, which queries Brave's own web index rather than reselling Google or Bing results. The main tool runs a web search with filters for country, language, freshness, safe search and result types, and can apply Goggles, Brave's custom re-ranking rules. Separate tools search news, images, videos, local businesses and points of interest around a location.

Brave's server also has two tools aimed at language models. The LLM context tool returns pre-extracted text from the top results, with limits on tokens, URLs and snippets, so the agent gets readable source material in one call. The summarizer tool turns a web search into an AI-written summary with optional inline references, using a key returned by a search made with summaries turned on.

Brave publishes the server on npm and as a Docker image, and it runs locally over stdio by default or over HTTP when you ask for it. You can allow or block single tools with environment variables. The GitHub repo had 1,478 stars on 2026-10-01; version 2 changed the default transport to stdio and stopped returning base64 image data.

Tools it gives your agent

• brave_web_search
• brave_local_search
• brave_place_search
• brave_image_search
• brave_video_search
• brave_news_search
• brave_summarizer
• brave_llm_context

When to use Brave Search

• You want search results from an index that is not Google's or Bing's.
• Your agent needs recent news on a topic, filtered to the last day or week.
• You need local businesses or places near a city or coordinates, with hours and ratings.
• You want short grounded text from search results without a separate scraping step.

When to pick something else

• Reading full pages or crawling a site: the server returns results and extracted snippets, not whole pages. Firecrawl covers that.
• Use without an API key: every call needs a Brave Search API key, and requests beyond the monthly free credit are billed.

What Brave Search needs

• A Brave Search API key
• Node.js or Docker to run the server locally

What the Brave Search MCP server costs

Free and open source under MIT; searches are billed to your Brave Search API account, which has a monthly free credit.

Current prices: brave.com ↗. We do not list prices: vendors change them often, so check the publisher's own page.

Other MCP servers for web search and scraping

• Apify: The Apify MCP server gives AI agents thousands of ready-made scrapers and automation tools from Apify Store, such as Google Maps, Instagram and search result scrapers, which the agent can find and run on demand.
• Exa: The Exa MCP server gives AI agents web search and page reading through Exa's search engine, plus optional advanced filters and Exa Agent runs for multi-step research with structured output.
• Firecrawl: The Firecrawl MCP server gives AI agents live web access through Firecrawl: search the web, scrape a URL into Markdown or JSON, map and crawl whole sites, and click through pages.
• Jina AI: The Jina AI MCP server gives AI agents Jina's Reader, Search, Embeddings and Reranker APIs: read any web page or PDF as Markdown, search the web, arXiv and SSRN, and rerank results.