Today's development of agent applications in general, but in particular related to society simulation, presents two aspects that are separated by a gap: agent methodologies and agent infrastructures. On the one hand, methodologies support the agent development in its first phases – mainly analysis and design – while infrastructures are clearly focused on the implementation and deployment. Unfortunately, there are few connections between them, leading to fragmented solutions. In this paper, we propose to bridge this gap by means of meta-models, and we show how the PASSI methodology and the RoleX infrastructure can be connected by following our approach.