The concept of role allows introducing organizational aspects in the development of multi-agent systems. In this chapter we elaborate the role-based model of computational multi-agent system and formalize this model in description logic. We show that the division of the ontology into axioms interpreted under the open- and closed-world assumption is necessary to express correctness conditions of the system. The ontology agent is proposed and implemented. This agent provides management services, e.g. run-time validation of the system or matchmaking of present groups and agents. An example of the simple computational multi-agent system is shown.