CUA

Active
GitHub HTML MIT

Description

CUA provides open-source infrastructure for Computer-Use Agents, including sandboxes, SDKs, and benchmarks to train and evaluate AI agents that control full desktops (macOS, Linux, Windows).

Key Features

  • Cua Driver: background computer-use agent without stealing cursor focus
  • Cross-platform sandboxes: unified API for macOS, Linux, Windows, Android
  • CuaBot: multi-agent computer-use sandbox CLI with native window display
  • Cua-Bench: OSWorld, ScreenSpot benchmarks and RL environments
  • Lume: near-native performance macOS virtualization on Apple Silicon
  • Integration with Claude Code, OpenClaw, Cursor and other major agent frameworks

Use Cases

💡 AI agents autonomously operating desktop applications for complex tasks
💡 Cross-platform GUI automation testing and regression verification
💡 Building reproducible computer-use agent benchmarks
💡 Running untrusted code or browser automation in secure sandboxes

Quick Start

Install Python package: pip install cua, create a sandbox with Sandbox.ephemeral(Image.linux()), control the desktop via sb.shell.run(), sb.screenshot(), sb.mouse.click() and other APIs.

Related Projects