This paper is concerned with the automated generalisation of vector geographic databases. It studies the possible synergies between three existing, complementary models of generalisation, all based on the multi-agent paradigm. These models are respectively well adapted for the generalisation of urban spaces (AGENT model), rural spaces (CARTACOm model) and background themes (GAEL model). In these models, the geographic objects are modelled as agents that apply generalisation algorithms to themselves, guided by cartographic constraints to satisfy. The differences between them particularly lie in their constraint modelling and their agent coordination model. Three complementary ways of combining these models are proposed: separate use on separate zones, “interlaced” sequential use on the same zone, and shared use of data internal to the models. The last one is further investigated and a partial re-engineering of the models is proposed.