Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops.
Spin up full, disposable OS environments where an agent can click, type, and navigate — safely contained.
A single SDK to wire your model into real screen + keyboard + mouse control across macOS, Linux, and Windows.
Reproducible tasks to measure whether your computer-use agent actually completes the work — not just demos.