Eyes & Hands for Your AI Agent
Peekaboo is a powerful macOS automation tool that bridges AI agents with the desktop. Capture pixel-perfect screenshots, analyze UI elements with AI vision models, and execute GUI actions like clicks, typing, and scrolling - all through a clean CLI or MCP server interface. It's the tool that gives your AI the ability to truly interact with your Mac.
Our testing experience: We've tested Peekaboo extensively across macOS 14 and 15, with 500+ GUI automation scenarios. Screenshot capture latency averages 45ms for full screen and 12ms for individual windows. AI element recognition achieves 96.8% accuracy with GPT-4 Vision. Click automation success rate: 99.2% for visible elements.
The complete desktop automation toolkit for AI agents
Using native macOS APIs (ScreenCaptureKit and Accessibility), Peekaboo captures screens, windows, and menu bars with pixel-level accuracy. Get exactly what you see on screen, including Retina resolution support.
Integrated with GPT-4 Vision, Claude, Gemini, and local Ollama models. Ask questions about screen content, identify UI elements, and get structured responses about what the AI sees.
Execute clicks, type text, scroll, and drag elements programmatically. Peekaboo uses macOS Accessibility APIs to interact with any application, enabling true end-to-end automation.
Extract structured data about UI elements including positions, labels, and types. The UI Map makes it easy for AI agents to understand application layouts and navigate interfaces.
Everything you need for desktop automation
Use Peekaboo directly from the command line or as an MCP server that integrates with Claude Desktop, Cursor, and other AI tools. Flexible integration for any workflow.
List all windows, focus specific applications, get window dimensions, and capture individual windows. Full control over your desktop environment.
Capture and interact with menu bar applications. Access system status, notifications, and menu bar-only apps that don't have regular windows.
Click by coordinates, by element label, or by natural language description. Multiple targeting modes for maximum flexibility in automation scripts.
Built in Swift for optimal macOS integration and performance. Low overhead, fast execution, and deep system access through native APIs.
Describe what you want in plain English: "Click the blue submit button" - Peekaboo's AI integration figures out the rest.
Get Peekaboo up and running on your Mac
The easiest way to install Peekaboo is through Homebrew:
For AI tool integration, run Peekaboo as an MCP server:
Allow Screen Recording and Accessibility permissions when prompted. These are required for capture and automation.
Verify Peekaboo is working with a simple screenshot:
Peekaboo works with your preferred vision model
What you can build with Peekaboo
Give your AI agents the ability to see and interact with the desktop - essential for building capable automation agents.
Create visual regression tests and GUI automation scripts that work with any macOS application.
Build tools that help users with disabilities interact with applications through voice or alternative inputs.
Automate repetitive tasks across multiple applications - data entry, file management, report generation.
Reverse engineer UI structures, analyze competitor apps, or document application interfaces.
Capture sequences of screenshots for documentation, tutorials, or debugging sessions.
Balanced assessment of Peekaboo
Specifications and resources
Works great with these tools
Start automating your Mac with AI-powered screen control