Mega-urbanization presents researchers with a network of densely interwoven problems that completely elude disciplinary boundaries. We report on the development of a spatial knowledge management and agent simulation framework that is designed to integrate closely with the process of trans-disciplinary research into the dynamics of complex human-environment systems. We argue that our choice of knowledge representation languages facilitates cross-domain collaboration and direct involvement of domain experts without prior experience in computer programming.In a run-through application example, we show how standardized knowledge engineering technologies are used to turn a conventional geodatabase into a self-documenting knowledge base that can flexibly interface with modern open-data infrastructures. The resulting cross-domain world model is then coupled to a graphical actor modeling language that specializes in the formulation of behavioral theories in terms of social roles, intentions, tasks, conditions and interaction. Finally, we describe how system theories expressed in this way are automatically translated into computer simulations.