Mcp 浏览器自动化

什么是 Algonius 浏览器？

Algonius 浏览器是一个开源的浏览器自动化服务器，旨在通过模型上下文协议（MCP）向外部 AI 系统暴露浏览器控制工具。这个创新工具允许开发者将浏览器功能集成到他们的应用程序中，从而实现无缝的自动化和网页交互控制。

Algonius 浏览器的特点

• 开源：Algonius 浏览器是完全开源的，允许开发者访问、修改和贡献其代码库。
• MCP 集成：它利用模型上下文协议促进 AI 系统与浏览器控制工具之间的通信。
• 安全：该浏览器在设计时考虑了安全性，确保自动化任务安全高效地执行。
• 用户友好的界面：该浏览器提供了一个简单明了的界面，方便开发者在没有复杂设置的情况下实现自动化任务。
• 社区支持：作为一个开源项目，它拥有一个不断壮大的贡献者和用户社区，提供支持和增强功能。

如何使用 Algonius 浏览器

1. 安装：首先从 GitHub 克隆代码库，并按照文档中提供的安装说明进行操作。
2. 设置：根据您的需求配置浏览器设置。这可能包括设置模型上下文协议以便与 AI 系统进行通信。
3. 自动化任务：利用提供的 API 创建自动化脚本，控制浏览器操作，如导航、表单提交和数据提取。
4. 测试：在受控环境中运行您的自动化脚本，以确保它们按预期功能运行，然后再在生产环境中部署。
5. 贡献：如果您有改进或功能要添加，请考虑通过在 GitHub 上提交拉取请求来为项目做出贡献。

常见问题解答

Q1: Algonius 浏览器是免费使用的吗？

A1: 是的，Algonius 浏览器是开源的，并且根据 Apache-2.0 许可证免费使用。

Q2: 我可以将 Algonius 浏览器与现有的 AI 系统集成吗？

A2: 当然可以！该浏览器设计用于通过模型上下文协议与外部 AI 系统协作，使集成变得简单。

Q3: 我在哪里可以找到 Algonius 浏览器的文档？

A3: 文档可在 GitHub 的代码库中找到，包括安装说明、API 参考和使用示例。

Q4: 我该如何报告问题或请求功能？

A4: 您可以通过在 GitHub 代码库中打开问题来报告问题或请求功能。社区会积极监控并回应反馈。

Q5: 是否有 Algonius 浏览器用户的社区？

A5: 是的，GitHub 上有一个活跃的用户和贡献者社区，您可以在这里提问、分享经验并协作项目。

<div align="center"> <img src="https://github.com/user-attachments/assets/1b2b1bc0-c7b4-4a45-83f5-4a6161831535" width="600" alt="Algonius Browser Banner" /> </div> <h1 align="center">MCP Browser Automation</h1>

🌐 Overview

Algonius Browser is an open-source MCP (Model Context Protocol) server that provides browser automation capabilities to external AI systems. It exposes a comprehensive set of browser control tools through the MCP protocol, enabling AI assistants and other tools to navigate websites, interact with DOM elements, and extract web content programmatically.

🎯 Key Features

• MCP Protocol Integration: Standard interface for AI systems to control browser automation
• Chrome Extension: Background service worker that handles browser interactions
• Native Messaging: Go-based MCP host that bridges Chrome extension with external tools
• Comprehensive Tool Set: 5 browser automation tools + 2 MCP resources
• Type Safety: Full TypeScript implementation with structured error handling
• Testing Suite: Comprehensive integration tests for all functionality

🛠️ Available MCP Tools

Navigation & Tabs

• navigate_to: Navigate to URLs with configurable timeout handling
• manage_tabs: Create, close, and switch between browser tabs

DOM Interaction

• get_dom_extra_elements: Advanced DOM element extraction with pagination and filtering
• click_element: Click DOM elements using CSS selectors or text matching
• set_value: Set values in input fields, textareas, and form elements
• scroll_page: Scroll pages up or down with customizable distances

📋 Available MCP Resources

Browser State Resources

• browser://current/state: Complete current browser state in AI-friendly Markdown format

  • Active tab information
  • All browser tabs with URLs, titles, and status
  • Real-time state updates via resource notifications

• browser://dom/state: Current DOM state overview in Markdown format

  • Page metadata (URL, title, scroll position)
  • First 20 interactive elements
  • Total element count with "more available" indicators
  • Simplified DOM structure
  • Auto-updates when page changes

🚀 Quick Start

1. Install Chrome Extension

From Chrome Web Store (Recommended):

<a href="https://chromewebstore.google.com/detail/algonius-browser-mcp/fmcmnpejjhphnfdaegmdmahkgaccghem" target="_blank"> <img src="https://github.com/user-attachments/assets/4c2c0b5e-8f63-4a8b-9a5e-2d7e8f3c9b1a" alt="Available in the Chrome Web Store" width="248" height="75"> </a>

1. Click the "Add to Chrome" button on the Chrome Web Store page
2. Confirm the installation when prompted
3. The extension will be automatically installed and ready to use

From Source (Development):

### Clone and build
git clone https://github.com/algonius/algonius-browser.git
cd algonius-browser
pnpm install
pnpm build

### Load in Chrome
### 1. Open chrome://extensions/
### 2. Enable "Developer mode"
### 3. Click "Load unpacked"
### 4. Select the 'dist' folder

⚠️ Important: The Chrome extension requires the MCP Host backend service to function properly. Please continue with step 2 to complete the installation.

2. Install MCP Host

One-Click Installation (Recommended):

Linux/macOS:

curl -fsSL https://raw.githubusercontent.com/algonius/algonius-browser/master/install-mcp-host.sh | bash

Windows (PowerShell):

iwr -useb https://raw.githubusercontent.com/algonius/algonius-browser/master/install-mcp-host.ps1 | iex

Manual Installation:

### Download latest release
wget https://github.com/algonius/algonius-browser/releases/latest/download/mcp-host-linux-x86_64.tar.gz

### Extract and install
tar -xzf mcp-host-linux-x86_64.tar.gz
cd mcp-host-linux-x86_64
./install.sh

3. Verify Installation

### Test the MCP host installation
mcp-host-go --version

### The MCP host will be automatically started when needed by the Chrome extension
### You should see the extension icon in your Chrome toolbar

✅ Success: Both components are now installed! The Chrome extension will automatically communicate with the MCP Host when browser automation is requested.

🔧 Integration Examples

Using with AI Assistants

Once installed, AI systems can use the browser automation tools and resources through the MCP protocol:

Tool Usage:

{
  "method": "tools/call",
  "params": {
    "name": "navigate_to",
    "arguments": {
      "url": "https://example.com",
      "timeout": 30000
    }
  }
}

Resource Access:

{
  "method": "resources/read",
  "params": {
    "uri": "browser://current/state"
  }
}

Common Workflows

Web Scraping:

1. navigate_to → Navigate to target site
2. Read browser://dom/state → Get page overview
3. get_dom_extra_elements → Get specific elements with pagination
4. click_element → Interact with elements
5. Read browser://dom/state → Extract updated content

Form Automation:

1. navigate_to → Go to form page
2. Read browser://dom/state → Identify form elements
3. set_value → Fill form fields
4. click_element → Submit form
5. Read browser://current/state → Verify completion

Multi-Tab Management:

1. Read browser://current/state → Check current tabs
2. manage_tabs → Create/switch tabs
3. navigate_to → Load content in each tab
4. Read browser://current/state → Monitor all tab states

Page Navigation with Scrolling:

1. navigate_to → Go to target page
2. Read browser://dom/state → Get initial page state
3. scroll_page → Scroll to load more content
4. get_dom_extra_elements → Extract newly loaded elements

🏗️ Architecture

External AI System
       ↓ (MCP Protocol)
   MCP Host (Go)
       ↓ (Native Messaging)
Chrome Extension
       ↓ (Chrome APIs)
    Browser Tabs

Components

• MCP Host: Go-based native messaging host that implements MCP protocol
• Chrome Extension: Background service worker with tool handlers
• Content Scripts: DOM interaction and data extraction utilities
• Integration Tests: Comprehensive test suite for all tools

🧪 Development

Build from Source

Prerequisites:

• Node.js 22.12.0+
• pnpm 9.15.1+
• Go 1.21+ (for MCP host)

Build Extension:

pnpm install
pnpm build

Build MCP Host:

cd mcp-host-go
make build

Run Tests:

### Extension tests
pnpm test

### MCP host tests  
cd mcp-host-go
make test

Development Mode

### Extension development
pnpm dev

### MCP host development
cd mcp-host-go
make dev

📊 Supported Platforms

MCP Host:

• Linux x86_64
• macOS Intel (x86_64) and Apple Silicon (arm64)
• Windows x86_64

Chrome Extension:

• Chrome/Chromium 88+
• Microsoft Edge 88+

📚 Documentation

Detailed documentation available in the docs/ directory:

• MCP Host Integration
• Click Element Tool
• Set Value Tool
• Navigate To Timeout
• Integration Testing

🤝 Contributing

We welcome contributions! Check out our CONTRIBUTING.md for guidelines.

Ways to contribute:

• Report bugs and feature requests
• Submit pull requests for improvements
• Add integration tests
• Improve documentation
• Share usage examples

🔒 Security

For security vulnerabilities, please create a GitHub Security Advisory rather than opening a public issue.

💬 Community

• Discord - Chat with developers and users
• GitHub Discussions - Share ideas and ask questions

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

👏 Acknowledgments

Built with these excellent open-source projects:

• Chrome Extension Boilerplate
• Model Context Protocol

Made with ❤️ by the Algonius Browser Team

Give us a star 🌟 if this project helps you build better browser automation!