Nowadays, agent-based modelling is more and more used to study complex socio-ecological systems. These last years have also seen the development of several agent-based simulation platforms. These platforms allow modelers to easily and quickly develop models with simple agents. However, socio-ecological systems need agents able to make decisions in order to represent human beings and the design of such complex agents is still an open issue: even with these platforms, designing agents able to make complex reasoning is a difficult task, in particular for modelers that have no programming skill. In order to answer the modeler needs concerning complex agent design, we propose a new agent architecture based on the BDI paradigm and integrated into a simulation platform (GAMA). This paradigm allows designing expressive and realistic agents, yet, it is rarely used in simulation context. A reason is that most agent architectures based on the BDI paradigm are complex to understand and to use by non-computer-scientists. Our agent architecture answers this problem by allowing modelers to define complex cognitive agents in a simple way. An application of our architecture on a model concerning forest fire and fire-fighter helicopters is presented.