AI Agent MCP Toolbox

🚀 A comprehensive MCP (Model Context Protocol) server that empowers AI agents with visual, development, and system monitoring capabilities.

✨ Features

• 📸 Cross-Platform Screenshots - Capture screens, windows, or regions on Linux/macOS/Windows
• 🔍 UI Element Inspection - Extract UI hierarchies with accessibility analysis
• ⚡ Performance Monitoring - Real-time system metrics and performance analysis
• 👁️ File System Watching - Monitor file/directory changes in real-time
• 🖱️ Screen Control - Click, type, scroll, and keyboard automation for AI agents
• 🎯 Claude Code Integration - Seamless integration with Claude Code IDE

🎯 Why This Matters

AI agents typically face constraints when working with visual interfaces and real-time system monitoring. This toolbox bridges that gap by providing:

• Visual Capabilities: Take screenshots and analyze UI elements
• Development Tools: Monitor file changes and system performance
• Accessibility Auditing: Check UI compliance and design patterns
• Cross-Platform Support: Works on Linux, macOS, and Windows

Perfect for AI-assisted development, UI/UX testing, system monitoring, and accessibility auditing.

🚀 Quick Start

Prerequisites

• Node.js 18+
• Claude Code (for AI agent integration)

Installation

Option 1: NPM Install (Recommended)

# Install globally
npm install -g @origindot./zigzag

# The package will be ready to use with Claude Code

Option 2: From Source

# Clone the repository
git clone https://github.com/Antonio476587/zigzag.git
cd zigzag

# Install dependencies
npm install

# Build the project
npm run build

# Set up platform dependencies
./setup-screen-control.sh  # Linux/macOS

Option 3: One-Line Installer

curl -sSL https://raw.githubusercontent.com/Antonio476587/zigzag/main/ai-agent-mcp-toolbox/install.sh | bash

Claude Code Configuration

For NPM Installation:

Add to your ~/.config/claude/mcp_settings.json:

{
  "mcpServers": {
    "ai-agent-toolbox": {
      "command": "npx",
      "args": ["@origindot./zigzag"],
      "env": {}
    }
  }
}

For Source Installation:

{
  "mcpServers": {
    "ai-agent-toolbox": {
      "command": "node",
      "args": ["/path/to/zigzag/dist/index.js"],
      "env": {}
    }
  }
}

🛠️ Available Tools

📸 Screenshot Tool

Capture screenshots using system commands with cross-platform compatibility.

Usage in Claude Code:

• "Take a screenshot of my screen"
• "Capture the current window"
• "Screenshot a specific region"

Parameters:

• target: 'screen' | 'window' | 'region'
• format: 'png' | 'jpg' | 'bmp' (default: 'png')
• window_title: Title of window to capture (for window target)
• region: {x, y, width, height} (for region target)

🔍 UI Inspect Tool

Analyze UI element hierarchies and check accessibility compliance.

Usage in Claude Code:

• "Inspect the UI of my browser"
• "Analyze accessibility issues in my app"
• "Extract UI element hierarchy"

Features:

• UI element tree extraction
• Accessibility issue detection
• Design pattern identification
• WCAG compliance checking

⚡ Performance Monitor Tool

Monitor system performance with detailed metrics and recommendations.

Usage in Claude Code:

• "Monitor my application's performance for 60 seconds"
• "Check system CPU and memory usage"
• "Analyze performance bottlenecks"

Metrics:

• CPU usage and trends
• Memory consumption
• GPU utilization (when available)
• Network I/O
• FPS monitoring
• Performance recommendations

👁️ File Watcher Tool

Monitor file and directory changes in real-time.

Usage in Claude Code:

• "Watch my src folder for JavaScript changes"
• "Monitor project files for modifications"
• "Set up file watching for hot reload"

Features:

• Multi-pattern file watching
• Event filtering (add, change, delete)
• Duration-based monitoring
• Status reporting

🖱️ Screen Control Tool

Automate mouse and keyboard interactions for AI agents.

Usage in Claude Code:

• "Click at position 100, 200"
• "Type 'Hello World'"
• "Press Ctrl+C"
• "Scroll down 3 times"
• "Drag from 100,100 to 200,200"

Actions:

• click, double_click, right_click: Mouse clicking
• type: Text input
• key: Keyboard shortcuts and special keys
• scroll: Scroll in any direction
• drag: Drag and drop operations
• move_mouse: Mouse positioning

Cross-Platform Support:

• Linux: Uses xdotool
• macOS: Uses cliclick
• Windows: Uses PowerShell automation

🧪 Testing

Run the included tests to verify functionality:

# Test individual tools
npm test

# Manual testing
node dist/index.js

📚 Examples

Basic Screenshot

echo '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"screenshot","arguments":{"target":"screen"}}}' | node dist/index.js

UI Inspection

echo '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"ui_inspect","arguments":{"target":"browser"}}}' | node dist/index.js

File Watching

echo '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"file_watcher","arguments":{"action":"start","path":"/path/to/watch","pattern":"*.js"}}}' | node dist/index.js

🏗️ Architecture

src/
├── index.ts          # MCP server entry point
├── server.ts         # Main server implementation  
├── types/            # TypeScript type definitions
├── tools/            # Individual tool implementations
│   ├── screenshot-system.ts
│   ├── ui-inspect-simple.ts
│   ├── performance-monitor-simple.ts
│   └── file-watcher.ts
└── utils/            # Utility functions

🌍 Platform Support

Platform	Screenshot	UI Inspect	Performance	File Watch	Screen Control
Linux	✅ Full	✅ Basic	✅ Full	✅ Full	✅ Full
macOS	✅ Full	✅ Basic	✅ Full	✅ Full	✅ Full
Windows	✅ Full	✅ Basic	✅ Full	✅ Full	✅ Full

Platform-Specific Notes

Linux:

• Screenshots: Uses system commands (import, gnome-screenshot, scrot, etc.)
• Screen Control: Requires xdotool (sudo apt install xdotool)

macOS:

• Screenshots: Built-in screencapture command
• Screen Control: Requires cliclick (brew install cliclick)

Windows:

• Screenshots: PowerShell with System.Drawing
• Screen Control: Built-in PowerShell automation

🤝 Contributing

Contributions are welcome! Please feel free to submit pull requests, report bugs, or suggest features.

Development Setup

# Clone and install
git clone https://github.com/Antonio476587/zigzag.git
cd zigzag
npm install

# Install platform dependencies
./setup-screen-control.sh  # Linux/macOS

# Development mode
npm run dev

# Build
npm run build

# Lint
npm run lint

# Test
npm test

📝 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

• Built on the Model Context Protocol
• Inspired by the need for AI agents to interact with visual interfaces
• Thanks to the Claude Code team for MCP integration

🆘 Troubleshooting

Screenshots Not Working

Linux: Install required screenshot tools:

# Ubuntu/Debian
sudo apt-get install imagemagick scrot

# Fedora
sudo dnf install ImageMagick scrot

macOS: No additional setup required

Windows: Ensure PowerShell execution policy allows scripts

Performance Monitoring Issues

Some metrics require elevated permissions. Run with appropriate privileges if needed.

File Watcher Not Responding

Check file permissions and ensure the target directory exists and is accessible.

📞 Support

For issues, questions, or feature requests, please:

1. Check the troubleshooting section above
2. Search existing issues on GitHub
3. Create a new issue with detailed information

Package Information

• npm: @origindot./zigzag
• GitHub: https://github.com/Antonio476587/zigzag
• Documentation: https://github.com/Antonio476587/zigzag#readme

---

Made with ❤️ for the AI development community