trycua/cua
Open-source infrastructure for building, training, and evaluating AI agents that can control full desktops across various operating systems.
Core Features
Quick Start
pip install cuaDetailed Introduction
Cua provides a robust open-source infrastructure designed for the development, training, and evaluation of AI agents capable of interacting with full desktop environments. It offers agent-ready sandboxes supporting various operating systems like macOS, Linux, Windows, and Android, accessible via a unified API. This enables agents to autonomously perform tasks by seeing screens, clicking buttons, and executing code, whether deployed in the cloud or locally using QEMU. Cua aims to simplify the creation of sophisticated computer-use agents by abstracting away OS-specific complexities and providing essential tools for agent development and benchmarking.