MCPRisingofficial publisher

Brave Search MCP

MCP : Standardises the connection between an AI application and systems that expose tools or data · requires tout client MCP

Independent web search for the agent via the Brave API.

A solid privacy-respecting option, but the free tier is gone.

GitHub · 1,527 ★

Overall score

8.5 /10

Security: nothing to flag

Utility

7.0

Reliability

10.0

Security

No caveat on record

Ease

5 min · Comfortable

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

AI agents need real-time web access to answer factual questions, monitor news, or find local businesses. Without a connection to a search engine, they remain limited to their training data, which is often outdated.

Brave Search MCP is an MCP server that connects your AI client to the Brave Search API. It exposes several distinct tools: general web search, news search, video search, image search, geolocated place search, an AI-optimized context tool, and an automatic summarization tool. The server runs in STDIO mode by default but can also be launched in HTTP mode.

This tool is aimed at developers and advanced users who configure AI agents in environments such as Claude Code or Cursor, and who want those agents to perform web searches without going through Google or Bing. Brave's search index is independent, making it a privacy-respecting option.

This server is not suited for users looking for a plug-and-play solution with no technical setup: you need to create an account on the Brave Search API platform, generate an API key, and manually integrate it into the MCP client configuration. Advanced local search and extra snippets require a paid Pro plan.

What it offers

·Comprehensive web search with filters for country, language, freshness, and content type.
·News search with freshness controls, defaulting to the last 24 hours.
·Video search with full metadata and thumbnail information.
·Image search with dimensions and confidence scores, without unnecessary base64 encoding since version 2.
·Geolocated place search with opening hours, contact details, ratings, and photos (paid).
·LLM Context tool that provides pre-extracted web content optimized for RAG pipelines and AI agent grounding.
·Automatic summarization of search results via Brave's summarization API, in two steps: web search with the summary option, then a call to the brave_summarizer tool.
·Extra snippets to enrich web and news search results (paid).

Our tips

To get an AI summary, first perform a web search with the summary parameter set to true, retrieve the returned key, then call the brave_summarizer tool with that key.
You can enable or disable specific tools using the BRAVE_MCP_ENABLED_TOOLS and BRAVE_MCP_DISABLED_TOOLS environment variables, which helps reduce session context by exposing only the tools you actually need.
If you deploy the server in a Docker container, set BRAVE_MCP_HOST to 0.0.0.0 so the server is reachable from outside the container, and restrict allowed origins using BRAVE_MCP_ALLOWED_ORIGINS.
The brave_llm_context tool lets you finely control the amount of context passed to the agent through the maximum_number_of_tokens and maximum_number_of_tokens_per_url parameters, which can reduce context window consumption.
For local searches without a Pro plan, the server automatically falls back to a standard web search: check the results to make sure they match your geographic intent.

Watch out

Advanced local search and extra snippets are reserved for paid Pro plans: without the appropriate subscription, these features degrade silently or are unavailable.
HTTP mode is unauthenticated by default: exposing the server on all network interfaces without configuring allowed origins and host restrictions is a real security risk.

In practice

Price: Freemium
Commercial use: Commercial use: yes
Available on: claude-code, cursor
Host: tout client MCP
Licence: MIT

Security and data

Rating: Security: nothing to flag
Your data: Data with the vendor
Maintenance: Active
Last release: 17 September 2026

Links and repository checked on 25 October 2026 · added on 1 November 2024

How to install Brave Search MCP

1. Create an account at https://brave.com/search/api/ and choose a Search or Answers plan. 2. Generate your API key from the Brave developer dashboard. 3. Store the key in the BRAVE_API_KEY environment variable or in a file referenced by BRAVE_API_KEY_FILE (recommended for Docker environments with mounted secrets). 4. Add the MCP server configuration to your client's configuration file (for example claude_desktop_config.json for Claude Desktop), pointing to the Brave Search MCP server. 5. Verify that the transport matches your environment: STDIO by default, HTTP if needed via the BRAVE_MCP_TRANSPORT variable. 6. Restrict exposed tools using BRAVE_MCP_ENABLED_TOOLS if you do not need all search types, to reduce the exposure surface.

node dist/index.js --transport stdio --brave-api-key <YOUR_API_KEY>

In the same category

Whole category Feed (memory, context) →