Brave Search MCP Server

@mikechao

About Brave Search MCP Server

An MCP Server implementation that integrates the Brave Search API, providing, Web Search, Local Points of Interest Search, Image Search, Video Search, News Search and LLM Context Search capabilities

#mcp #image-search #news-search #video-search #web-search #brave-search #model-context-protocol

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "brave-search-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "brave-search-mcp"
      ],
      "env": {
        "BRAVE_API_KEY": "your_key_here"
      }
    }
  }
}

Tools

Execute web searches using Brave's API

Get images from the web relevant to the query

Searches the web for news

Search for local businesses, services and points of interest

Search the web for videos

Pre-extracted web content optimized for AI agents, LLM grounding, and RAG pipelines.

Overview

What is Brave Search MCP Server?

The Brave Search MCP Server integrates the Brave Search API, providing web, image, news, video, local points of interest, and LLM context search capabilities. It is designed for developers and AI agents who need structured search results within the Model Context Protocol (MCP) ecosystem.

How to use Brave Search MCP Server?

Obtain a Brave Search API key from the Brave developer dashboard, then run BRAVE_API_KEY="your_key_here" npx -y brave-search-mcp for stdio mode or add --http for Streamable HTTP mode. Configuration can be done via environment variables or a TOML file (set BRAVE_MCP_CONFIG). To use with ChatGPT, run in HTTP+UI mode, expose via ngrok, and add a connector in ChatGPT settings.

Key features of Brave Search MCP Server

Web, image, news, and video search
Local points of interest search
LLM context search with customizable token and snippet limits
Streamable HTTP and stdio transport modes
Configurable request guardrails, auditing, and authentication
UI mode for use with ChatGPT web interface

Use cases of Brave Search MCP Server

Grounding AI agent responses with real-time web search results
Enabling RAG pipelines with pre-extracted web content
Searching for local businesses, restaurants, or services
Discovering recent news with freshness filtering
Fetching video and image results for multimedia applications

FAQ from Brave Search MCP Server

What API key is required?

You must have a Brave Search API key. Sign up at the Brave Search API website, choose a plan, and generate your key from the developer dashboard.

What runtime modes are supported?

The server supports stdio mode (default) and Streamable HTTP mode (enable with --http). HTTP mode listens on port 3001 by default.

How do I configure feature settings?

Feature settings can be set via environment variables (Env mode) or a TOML configuration file (File mode, set BRAVE_MCP_CONFIG). The API key, port, and host are always read from the environment.

Can I use this server with ChatGPT?

Yes. Run the server with --http --ui, create a local tunnel with ngrok, and add the server as an MCP Connector in ChatGPT's developer mode. Authentication is set to "No Auth" for this method.

What are the search result limits?

Web, news, video, and local searches have a maximum count of 20 results (default 10 or 5 for local). Image search max count is 50. LLM context search has configurable limits on tokens (max 32768), snippets (max 100), and URLs (max 50).

Frequently asked questions

What API key is required?

You must have a Brave Search API key. Sign up at the Brave Search API website, choose a plan, and generate your key from the developer dashboard.

What runtime modes are supported?

The server supports stdio mode (default) and Streamable HTTP mode (enable with `--http`). HTTP mode listens on port 3001 by default.

How do I configure feature settings?

Feature settings can be set via environment variables (Env mode) or a TOML configuration file (File mode, set `BRAVE_MCP_CONFIG`). The API key, port, and host are always read from the environment.

Can I use this server with ChatGPT?

Yes. Run the server with `--http --ui`, create a local tunnel with ngrok, and add the server as an MCP Connector in ChatGPT's developer mode. Authentication is set to "No Auth" for this method.

What are the search result limits?

Comments

Basic information

More Search MCP servers

Exa MCP Server 🔍

exa-labs

Exa MCP for web search and web crawling!

4.7K

Google News MCP Server

ChanMeng666

【Star-crossed coders unite!⭐️】Model Context Protocol (MCP) server implementation providing Google News search capabilities via SerpAPI, with automatic news categorization and multi-language support.

124