Pairs metric-scale 3D scans of real environments with egocentric task recordings, estimated body and hand motion, and hierarchical annotations for physical AI research. The first release contains 126 scenes; the dataset is licensed CC BY-NC 4.0.