Within the last decades, the design of more sustainable cities has emerged as a central society issue. A city, in the early stage of its design process, is modeled as a balanced set of urban shapes (residential, commercial, or industrial units, together with infrastructures, schools, parks) that need to be spatially organized following complex rules. To assist urban planners and decision makers in this largely manual and iterative endeavor, we propose the design of a computer-aided decision tool which first automatically organizes urban shapes over a given empty territory, and then offer interactive manipulators that allow the experts to modify the spatial organization, while maintaining relations between shapes and informing experts of the impact of their choices. We cast the problem as a Local Search optimization in which we perform a sequence of swaps between urban shapes, starting from a random initial assignment. We extend the algorithm with novel heuristics to improve computational costs and propose an efficient distributed version. The same algorithm is used for the automated and interactive stages of the design process. The benefits of our approach are highlighted by examples and feedbacks from experts in the domain.