Brave Browser MCP Server
Gives AI agents full control over Brave Browser, optimized for local LLMs to ensure low latency and high reliability.
README
🦁 Brave Browser MCP Server
A powerful Model Context Protocol (MCP) server that gives AI agents full control over your Brave Browser. Unlike standard automation, this server is optimized for local LLMs (like Qwen3.5 27B) to ensure low latency and high reliability.
⚡ Quick Setup
1. Install dependencies
pip install mcp[cli] playwright
playwright install chromium
Note: Even though we use Brave, Playwright needs its driver components to control it.
2. Add to Claude Desktop (or your MCP Host)
Edit your claude_desktop_config.json:
{
"mcpServers": {
"brave-browser": {
"command": "python",
"args": ["C:/PATH/TO/MCP-Brave-Browser/server.py"]
}
}
}
🔧 Optimized Tools (27)
Speed & Discovery (VITAL for Local Models)
| Tool | Description |
|---|---|
page_summary |
Instant Orientation. Returns Title, H1s, and Meta info. Use this first to save context. |
find_elements |
Targeted Search. Finds specific items by text/URL. Faster than reading a whole page. |
snapshot |
Accessibility Tree. Returns elements with ref=N IDs. |
Navigation
| Tool | Description |
|---|---|
navigate |
Opens a URL. |
get_page_info |
Returns current URL and page title. |
tabs |
List, open, close, or switch tabs. |
Page Interaction
| Tool | Description |
|---|---|
click / hover |
Interact via CSS selector or ref=N from snapshot. |
type_text |
Real-time character typing. |
fill_input |
Instant form filling. |
take_screenshot |
Visual verification (Timestamped in /screenshots). |
🚀 Performance Tips (Context & Compute)
If you are using a local 4B-9B model, full page snapshots can be slow. Use these strategies to make it lightning fast:
- Filter Noise: Use
snapshot(interactive_only=True). This strips outdiv,section, andnavcontainers, returning only the interactive elements. It usually reduces context size by 70%. - Scope your Search: Use
snapshot(selector="#main-content")to only see a specific part of the page. - Use Summaries First: Instead of taking a snapshot immediately, call
page_summary()to understand where you are. - Targeted Discovery: Use
find_elements(query="Login")if you just need to find one specific button.
⚙️ Configuration
| Setting | Default | Override |
|---|---|---|
| Brave path | C:\Program Files\BraveSoftware\Brave-Browser\Application\brave.exe |
Set BRAVE_EXECUTABLE_PATH env var |
| Profile Path | ./brave_mcp_profile |
Managed automatically (Persistent sessions) |
📦 Dependencies
mcp[cli]— FastMCP frameworkplaywright— High-speed browser automation engine
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。