Summary form only given. This paper proposes a modeling approach easy to master in practice and from which one could directly derive a formal model capturing the main features of the dynamic interactions involved in multiagent systems (MAS). This approach follows two steps. The first consists to describe agent interactions in a MAS using Agent Unifying Model Language (AUML). The second step consists in translating this AUML description into a recursive colored Petri net (RCPN) model by following suitable rules. The choice of the RCPN formalism is motivated by its ability to model the dynamic agent planning ensuring the interleaving between planning and execution of complex actions and the availability of formal analysis and verification methods and tools.