DeepYardDeepYard
C

CUA

Open-source infrastructure for Computer-Use Agents with sandboxes and benchmarks — 13K+ stars

Open SourceFree

About

CUA (Computer-Use Agents) is open-source infrastructure for building, training, and evaluating AI agents that control full desktops. Includes CuaBot (multi-agent sandbox CLI with native window rendering, H.265 video, shared clipboard), Cua (agentic UI automation framework with isolated code execution), Cua-Bench (evaluation platform for OSWorld, ScreenSpot, Windows Arena benchmarks), and Lume (macOS/Linux virtualization with near-native Apple Silicon performance). Agents see screens, click buttons, and complete tasks autonomously across macOS, Linux, and Windows.

Details

Typesandboxing, benchmarking, virtualization, computer-use
Integrationsmacos, linux, windows, docker
Languagepython, typescript, swift

Tags

computer-usesandboxingbenchmarksvirtualizationdesktop-automationpython