Abstract : The preliminary design aircraft can be viewed as a multi-disciplinary, multi-objective optimisation problem which consists in finding the values of the aircraft description parameters and of its performances. These values are calculated with disciplinary knowledge. The interdependencies between the parameters and non linearities lead to complex calculus which cannot be realized efficiently by classicmethods. We propose an approach based on adaptive multiagent systems in which agents represent the disciplines, the design parameters and the aircraft performances. In using disciplinary knowledge and a cooperative behaviour, the agents collectively reach the parameters values satisfying constraints and performances. We show this approach affords interesting capabilities such as sensibilities analysis, Pareto front or dynamic interactions.