Manufacturing control systems typically include layers of logic and heuristics making them difficult to develop. Furthermore, in agile manufacturing planning and control (MPC) environment, the dynamics of decision-making interactions between agile MPC system components are poorly understood and can be undependable because of the absence of a master controller or optimizer. This paper addresses these challenges by developing a multi-layer architecture for a supervisory decision logic unit (DLU) of an agile MPC system. After presenting a dynamic model for the system, a decision logic unit for that system that links the operational level with the higher enterprise level strategy is described. The architecture of the DLU is composed of three layers where the first layer is responsible for dynamically managing the selection of the different MPC policies that suits the market strategy. The second layer describes an algorithm for optimal parameters settings for each of the MPC policies. Finally the third layer of the architecture is responsible for the automatic on-line control of manufacturing system to maintain required production, work-in- process and inventory levels.