openclaw / Peekaboo
Gives AI agents eyes and hands on macOS through a unified CLI, MCP server, and menu-bar app with native accessibility inspection.
活跃维护 MIT Swift Tracked
5.4k 413 2 小时前
CINPM
ai macos mcp screenshots swift
星标趋势
数据积累中,暂无足够数据生成趋势图
AI 分析
项目摘要
Peekaboo is a macOS CLI and optional MCP server that enables AI agents to capture screenshots, inspect accessibility structures, and automate UI interactions. It combines screen capture, structured UI mapping, and visual question answering through local or remote AI models into a unified tool.
为什么值得关注
Peekaboo solves a critical gap in AI agent capabilities by giving them visual perception and UI control on macOS, with native accessibility inspection and a clean MCP interface that integrates seamlessly with popular AI coding clients like Cursor and Claude Code.
优势
- Unified CLI, MCP server, and menu-bar app with multiple installation methods (Homebrew, npm, signed DMG)
- Active development with 37 releases in 6 months and 42 contributors
- Native macOS accessibility inspection with structured UI element mapping and opaque IDs
- Optional VQA integration with local or remote AI models for screenshot analysis
- Well-structured documentation with clear permission guides and quick start
局限性
- macOS 15+ only with no cross-platform or Linux/Windows support
- Requires manual permission setup for Screen Recording and Accessibility
- No Docker containerization or headless server execution option
使用场景
- AI agent automation of desktop workflows and GUI testing
- Visual question answering on captured screenshots and application states
- Accessibility auditing and structured UI inspection for testing
- Building MCP-connected AI tools that interact with native macOS applications
目标用户: Developers building AI-powered macOS automation, AI agent developers needing visual/UI capabilities, and macOS power users creating AI-assisted workflows
学习曲线: 中
分析模型:LongCat-2.0 | 分析时间:13 天前