This paper presents a real-time cooperative multi-target tracking system. The system consists of a group of Active Vision Agents (AVAs, in short) representing a logical model of a network-connected computer with an active camera. All AVAs track their target objects cooperatively by interacting dynamically with each other. As a result, the system as a whole can track multiple moving objects simultaneously under complicated dynamic situations in the real world. To implement real-time cooperation among AVAs, we designed a three-layered interaction architecture. In each layer, parallel processes mutually exchange a variety of information for effective cooperation. We employed the dynamic memory architecture to achieve real-time information exchange. Experimental results demonstrated that AVAs track their target objects cooperatively in real-time while adaptively changing their roles.