In this paper, we argue that a shift is needed for plan-based architectures to support versatile, long-lived systems. Our hypothesis is that we need to integrate plan generation, plan analysis and plan adaptation within a wholesome framework that allow a seamless integration of planning and plan execution activities, i.e. that the architectural focus should move away from planning towards so-called plan management. We present, in this paper, a software framework for plan management, which is based on a novel plan representation. This representation has been designed so that the actual context of the execution is available at all times: what the robot is doing and–more importantly–why it is doing it. Our plan manager implementation is available as open source, and has already been used on three different live systems, on which it demonstrated its capabilities.