Outdoor Ambient Assisted Living at a metropolitan area level, requires to develop highly scalable systems that are tolerant with respect to many different types of faults. The software components that are immersed in the environment and that form the fixed infrastructural part of the system, have to be strongly coupled with all the mobile and volatile components in order to obtain collective cooperative processes that let the environment to transparently show some degree of intelligent behavior. In this paper, we aim at presenting a system for the intelligent management of the parking requests that come in a metropolitan area. In order to deal with all issues that arise from mobile devices and to emphasize software partition as a key methodology in AmI, we designed the parking system using the agent paradigm. Due to the general organization of a metropolitan area, we introduced a hierarchical model to better deal with all the details at different levels of abstractions. The agent-based software structure reflects the topographic hierarchy: the system workflow is then detailed using cooperative processes between agents, that are build on protocolled interactions that are based on message exchanges and agent mobility. A first prototype has been developed. The infrastructural part has been simulated andreal mobile device, namely HTC-Android smartphones, have been used. JADE and its plug-in for mobile device JADE-LEAP, has been used as the platform for agents.