This paper introduces a programming environment and architecture for the development of agent based cooperative applications using a role based approach. We focus on the cooperative aspects by introducing cooperation processes (CP) as a concept of its own. CPs describe all and only the coordination and cooperation parts of an application. The explicit documentation of the coordination and cooperation mechanisms used in a MAS allows their evaluation and reuse. We are able to change existing and introduce new cooperation processes at runtime without modifying the existing agents. We specify the cooperative behaviour of an agent in a separate role description. The interconnection of these roles constitutes the CP. Describing cooperation independent from concrete agents allows to build heterogeneous, federative and transformable MAS. We show how agents decide what roles to accept and how the agent-role interaction works. Finally we present the ROPE framework and runtime environment.