Multi-agent systems represent the most recent technology to design and develop distributed applications. Agents can also gain advantage from mobility, i.e., the capability of changing execution environment in world modeled as a set of nodes with different resources and services. This technology calls for appropriate local infrastructures to manage the interactions among agents and between agents and environments. We propose to build such infrastructures on the basis of roles, which represent the behavior and the capabilities of agents. Some examples are reported to explain the exposed concepts in concrete fields.