While agents have emphasised the notion of active software components, they are not likely to be the only active components in agent-based systems. In this paper, we first discuss the general notion of active system, and show how it relates with the issue of the consistent observation of distributed and heterogeneous multi-component systems. Then, we introduce the concept of boundary interface as a methodological abstraction for the engineering of active systems, which allow observer agents to be provided with three different kind of consistent system views, featuring observable, controlled, and constrained consistency, respectively.