Digital twins enable real-time, model- and databased simulations of physical assets or systems yielding more effective monitoring, analysis, and decision-making for enhanced performance and efficiency. The construction industry is highly interested in utilizing digital twins due to their potential to facilitate collaboration along the building life cycle, optimizing building energy usage, asset life cycle management, and data-driven decision-making to achieve sustainability objectives. Currently, necessary technological solutions that effectively tackle the inherent challenge of federating and integrating models and data that arises in these scenarios are however missing. Expanding upon the achievements of model-based engineering and its technological infrastructure, we propose a model repository for digital twins that tackles the models-meet-data challenge and establishes a technological basis for collaborative workflows. To this end, we carried out an extended requirements elicitation, architectural synthesis, and validation procedures. Preliminary empirical findings using an initial prototype suggest that the repository exhibits the necessary scalability, performance and isolation for collaborative work.