The growth in broadband access and increases in terminal capability have enabled us to provide large-scale multimedia services. The traditional unicast architecture cannot support such a service because the transportation from a content server to many end hosts causes the overload to a content server. Overlay multicast techniques have been proposed to solve this problem. This high-layer technology allows us to provide large-scale multicast services even for mobile networks. Also, the recent rapid growth of mobile broadband brought the large-scale mobile multicast services into reality. However, stream relays by users are indispensable in overlay multicast. Moreover, in mobile environments, users' willingness to relay depends on user states such as remaining battery charge, movement conditions, and the relationship with the relay requestor. In this paper, we investigate user actions and states in mobile overlay networks and model user actions as agents. Then, we discuss how user actions affect the performance of the mobile overlay network and which user states we should pay attention to in order to build effective networks through multi-agent simulations.