Abstract : Decision Support Systems (DSS) should help users to take the best decisions as earlier as possible. In this thesis, we are interested in such DSS which are based on the multiagent paradigm. Indeed, multiagent systems are often the best way used to build computer systems that need multicriteria queries because they permit to design complex computerized systems. Moreover, in DSS, temporal constraints are very strong and in a lot of cases partial or incomplete results obtained early are preferred and more useful than complete and accurate results obtained too late. Anytime techniques seem to be a good solution in order to deal with these situations, that is to take into account real-time aspects in this framework. In this document, we present a method and a model to design a real-time multiagent system is based on anytime techniques. Furthermore, we have taken into account the inherent distributed aspect of current DSS in our model.