Open systems are characterized by heterogeneous participants which can enter or leave the system at will. Typical examples are e-commerce applications or information agent systems. Agents (e.g. personal assistants for buying things on the Internet) will only temporarily take up roles (e.g. a buyer in on-line auctions). This creates the need to define precisely what it means that an agent ”takes up” a role and ”enacts” it. In this paper we present ongoing research on the determination of the conditions under which an agent can enact a role and what it means for the agent to enact a role. We define possible relations between roles and agents and discuss architectural and functional changes that an agent must undergo when it enters an open agent system. Categories and Subject Descriptors I.2.11 [Artificial Intelligence]: Distributed Artificial Intelligence—multi-agent systems; D.2.10 [Software Engineering]: Design—methodologies General Terms Design, Languages, Theory Keyw...