lahfir / agent-desktop

Rust-based desktop automation that gives AI agents reliable computer use through OS accessibility trees instead of pixel guessing

活跃维护 Apache 2.0 Rust Tracked
1.8k 129 21 小时前
CI
accessibility accessibility-api ai-agents automation cli desktop-automation macos mcp rust computer-use computer-use-agent computer-use-agents
GitHub 对比

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

Agent Desktop is a Rust-based tool that enables AI agents to reliably interact with desktop applications through OS accessibility trees rather than pixel-based screenshots. It provides stable UI element references and safe, retryable actions, with a CLI and C-ABI for multi-language integration.

为什么值得关注

Unlike most computer-use approaches that rely on screenshots and pixel guessing, Agent Desktop leverages OS accessibility APIs to provide stable, retryable UI interactions — a more reliable paradigm for AI agent desktop automation.

优势

  • Accessibility-tree approach provides stable, retryable actions vs pixel-based guessing
  • Rust implementation offers performance and safety with single-binary deployment
  • C-ABI enables easy integration with Python, Swift, Go, Ruby, Node, and C
  • Active development with 29 releases in 6 months and CI/CD pipeline

局限性

  • Primarily macOS-focused based on topics, potentially limited cross-platform support
  • No Docker support or examples provided for quick onboarding
  • Relatively new project with 23 open issues still being resolved

使用场景

  • AI agent desktop automation and computer use
  • Reliable UI-driven task automation for agent systems
  • Multi-application workflow orchestration by AI agents
  • Accessibility-based UI testing and interaction
目标用户： AI agent developers and researchers building computer-use capabilities, and automation engineers creating agent-driven desktop workflows
学习曲线： 中
分析模型：LongCat-2.0 | 分析时间：27 天前