Advances in smart grid technology have yet to coalesce into a comprehensive solution integrating the landscape of future power systems. The microgrid concept may offer a solution for combining advanced components and enabling technologies within an infrastructure that must expand to meet emerging demands. As autonomous power system entities, microgrids require robust real-time power management and control to simultaneously operate jointly with the utility, provide reliable service, and help achieve customer-driven objectives utilizing local power system assets. In this paper, a decentralized control architecture for microgrids is presented, along with a simulation environment appropriate for on-going investigations into real-time, agent-based decision-making. Challenges faced by operating self-organizing multi-agent system (MAS) are presented, as well as results for a representative power management scenario for a multi-asset microgrid capable of operating in grid-interconnected or islanded mode. The system and formulations presented demonstrate the viability and capability of decentralized agent-based control for microgrids and illustrate their potential towards achieving smart grid goals.