Urban planning and design have a considerable impact on the economic performance of cities and on the quality of life of the population. Efficiency at this level is hampered by the lack of integrated instruments for formulating, generating, and evaluating urban plans. This chapter describes the theoretical foundations of a research project, called City Induction, aimed at the creation of a model for the development of such an instrument, departing from existing theories, which are integrated through a discursive grammar. The proposed model is composed of three sub-models: (1) a model for formulating urban programs from the analysis and interpretation of the context, based on Alexander's pattern language; (2) a model for generating urban plans that match the program, based on Stiny's shape and description grammars; and (3) a model for evaluating urban plans, that can be used for analyzing, comparing and ranking alternative solutions, departing from Hillier's space syntax. A common urban space ontology guarantees the syntactic and semantic interoperability among the three sub-models. This ontology will be used to structure and codify information into a Geographic Information System (GIS), which will be the kernel for the computer implementation of the larger model. A CAD system is used to construct 3D models from contextual information stored in the GIS. In short, following Stiny and March's design machines concept, the goal is to create an urban design machine that is able to produce flexible urban plans at the site planning level.