cua

Cua is an open-source infrastructure for computer-use agents, providing sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops across Linux, Windows, macOS, and Android. It allows for scaling computer fleets, running training, evaluation, and data-generation workloads in parallel, and provides a unified API for cross-OS fleets. Cua's features include snapshot-native rollouts, batch throughput, and verified data.

cua is best for best for developers and researchers working with computer-use agents and ai models that require desktop automation. It sits in the Agents category and is a freemium option.

cua is listed as freemium. Check the official website for current, detailed pricing tiers.

cua scored 8.2 out of 10 in 2026, based on six weighted metrics: usefulness, quality, ease of use, value, reliability, and popularity.

Supports multiple operating systems, including Linux, Windows, macOS, and Android. Provides a unified API for cross-OS fleets. Offers snapshot-native rollouts and batch throughput for efficient testing and evaluation.

May require significant technical expertise to set up and use. Limited documentation and support resources available. Potential security risks associated with running computer-use agents on desktop environments.

cua is a solid, well-rounded option for best for developers and researchers working with computer-use agents and ai models that require desktop automation. Its 8.2/10 score reflects dependable performance in the Agents category.