
用于谷歌搜索的模型上下文协议服务器。一款基于Playwright的模型上下文协议(MCP)工具,可绕过搜索引擎的反机器人机制,执行谷歌搜索并提取结果,...
概览
用于谷歌搜索的模型上下文协议服务器。基于Playwright的模型上下文协议(MCP)工具,绕过搜索引擎的反机器人机制,执行谷歌搜索并提取结果,为Claude和Cursor等AI助手提供实时搜索功能。 - :使用浏览器指纹伪造和真实用户行为模拟以避免检测 - :遇到验证码时切换为标题模式,允许用户完成验证 - 保存浏览器会话状态以减少重复验证的需求 - :使用多种选择器组合以适应谷歌搜索页面的变化 - :实现模型上下文协议,便于与AI助手集成 - 支持不同语言和地区的搜索结果 1. 进入MCP>光标设置 2.点击 + 添加新MCP服务器 3。填写表格:- 名称:google-search(或您喜欢的任何名称)- 类型:命令 - 命令:npx -y @mcp-server/google-search-mcp@latest 添加一个 .cursor/mcp。
概述(AI 翻译)
用于谷歌搜索的模型上下文协议服务器。 一款基于Playwright的模型上下文协议(MCP)工具,绕过搜索引擎的反机器人机制,执行谷歌搜索并提取结果,为Claude和Cursor等AI助手提供实时搜索功能。
特色
- 反机器人绕过:利用浏览器指纹伪造和真实用户行为模拟来规避检测
- 自动验证码处理:遇到验证码时切换到头置模式,允许用户完成验证
- 状态持久性:保存浏览器会话状态,减少重复验证的需求
- 适应性:使用多种选择器组合来适应谷歌搜索页面的变化
- MCP 集成:实现模型上下文协议,便于与 AI 助手集成
- 多语言支持:支持不同语言和地区的搜索结果
使用光标
安装——全球
使用 npx 运行 MCP 服务器:
在你的光标IDE中
- 前往
Cursor Settings>MCP - 点击
+ Add New MCP Server - 填写表格:
- 名称:
google-search(或你喜欢的任何名字) - 类型:
command
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
Google Search MCP
Model Context Protocol server for google search. A Playwright-based Model Context Protocol (MCP) tool that bypasses search engine anti-bot mechanisms, performs Google searches, and extracts results, providing real-time search capabilities for AI assistants like Claude and Cursor.
Features
- Anti-Bot Bypass: Uses browser fingerprint spoofing and real user behavior simulation to avoid detection
- Automatic CAPTCHA Handling: Switches to headed mode when encountering CAPTCHAs, allowing users to complete verification
- State Persistence: Saves browser session state to reduce the need for repeated verification
- Adaptability: Uses multiple selector combinations to adapt to changes in Google search pages
- MCP Integration: Implements the Model Context Protocol for easy integration with AI assistants
- Multi-language Support: Supports search results in different languages and regions
Using with Cursor
Installation - Globally
Run the MCP server using npx:
npx -y @mcp-server/google-search-mcp@latest
In your Cursor IDE
- Go to
Cursor Settings>MCP - Click
+ Add New MCP Server - Fill in the form:
- Name:
google-search(or any name you prefer) - Type:
command - Command:
npx -y @mcp-server/google-search-mcp@latest
- Name:
Installation - Project-specific
Add an .cursor/mcp.json file to your project:
{
"mcpServers": {
"google-search": {
"command": "npx",
"args": [
"-y",
"@mcp-server/google-search-mcp@latest"
]
}
}
}
Development
yarn install
Build the project
yarn build
Usage
Running as an MCP Server
yarn start
Using with MCP Inspector
To debug the server, you can use the MCP Inspector:
# First build the project
yarn build
# Start the MCP Inspector and server
npx @modelcontextprotocol/inspector node dist/index.js
Parameters
The search tool accepts the following parameters:
query(required): Search query stringlimit(optional): Number of search results to return, default is 10timeout(optional): Search operation timeout in milliseconds, default is 60000language(optional): Language for search results, e.g., zh-CN, en-US, default is zh-CNregion(optional): Region for search results, e.g., cn, com, co.jp, default is cn
How It Works
- The tool uses Playwright to control a Chromium browser to perform Google searches
- It avoids bot detection through browser fingerprint spoofing and real user behavior simulation
- When encountering CAPTCHA verification, it automatically switches to headed mode for user completion
- It extracts search results and returns them in a structured format
- It saves browser state for reuse in subsequent searches
Advanced Configuration
Browser State File
安装
npx -y @mcp-server/google-search-mcp@latest配置
{
"mcpServers": {
"google-search": {
"command": "npx",
"args": [
"-y",
"@mcp-server/google-search-mcp@latest"
]
}
}
}