This paper presents a model-based graphical tool for supporting multiscale modeling of software architectures. The implemented solution is based on a rule-oriented iterative modeling process ensuring transition from a coarse-grained description to a fine-grained description. The rules manage the refinement between scales and are applied with respect to the system constraints. Multi-scale architectures are modeled graphically using UML notations. To ensure model consistency, our approach supports model transformation and validation of UML models with OCL constraints. In order to experiment our approach, we tested it with a predictive and preventive system dedicated to the smart home application for maintaining personalized medicine at home. This system is helpful for people with loss of autonomy, exposed to risks of accidents or needing a precise daily medical follow-up. We implement the resulting models using the Eclipse platform.