Peekaboo

Peekaboo is a macOS automation toolkit and MCP server that enables AI agents to capture screen pixels, read accessibility trees, and control desktop inputs.

Peekaboo functions as a macOS automation toolkit designed for both human operators and AI agents. It captures pixel-accurate screen content, reads the accessibility tree, and executes input commands like clicking, typing, scrolling, and dragging. The toolkit includes a CLI, an MCP server, and an agent runtime, allowing tools like Codex, Claude Code, and Cursor to perform desktop tasks. Users can run natural-language plan-and-act loops with support for multiple providers including OpenAI, Claude, Grok, Gemini, and Ollama. The system uses a daemon-based architecture to manage lifecycle, socket communication, and TCC permissions.

Peekaboo at a glance

Best for: Software engineer
Runs on: macOS, CLI
Pricing: Free
Released: May 2025
Category: AI agent tool
Website: peekaboo.boo

Key features

Pixel-accurate screen, window, and menu-bar capture
Accessibility tree reading for UI context
Desktop input execution including clicks, typing, and drag-and-drop
MCP server to expose tools to AI clients
Agent runtime for natural-language planning and action loops
Support for OpenAI, Claude, Grok, Gemini, and Ollama providers
Daemon-based runtime with warm start and socket support

Use it for

Automating desktop applications via natural language instructions
Connecting AI agents like Claude Code to macOS desktop controls
Performing UI testing and interaction through CLI commands
Capturing annotated accessibility maps for visual AI tasks

Alternatives to Peekaboo

BrowserOS — Runs entirely locally and lets AI agents use real user logins to complete tasks.
Mac Computer Use — Enables local desktop automation through native Mac windows and Accessibility controls.
Coucou — It provides open-source, real-time agent monitoring and approvals directly from the screen notch.
SupaYatta — Tracks Claude Code, Cowork, and Codex sessions in the Mac notch with instant alerts.
Yomado — Enables visual agents to work in desktop apps without requiring app-specific MCPs.
Windows-MCP — Enables any LLM to perform native Windows tasks without complex vision models.
Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
Offscreen — Runs agent tasks in fresh VMs so your desktop stays free for your own work.
OpenViking — Unifies agent memory, knowledge RAG, and skills into a single directory structure.
Point and Yap — Connects screen selections and voice instructions directly to your coding agent.
agent-device — Token-efficient snapshots and semantic references for native mobile and TV apps.
Alook — Connects existing local coding agents to team communication channels.

All Peekaboo alternatives