We present a framework and algorithms for multiresolution modeling of point-sampled geometry. Most of our techniques extend the key ingredients of recent multiresolution editing concepts, including geometry smoothing, simplification, and offset computation, to point clouds. Using these building blocks, we designed a powerful and versatile hierarchical representation for point-based modeling. An intuitive editing metaphor along with dynamic resampling techniques facilitates modeling operations at different scales and enables us to effectively handle very large model deformations. In our approach all processing is performed directly on point-sampled geometry without the need for intermediate local triangulation or meshing.