In a network environment, it is the interaction and cooperation among distributed entities that make the computation dynamic and alive. But until recently, this interaction and cooperation still relied on expensive, complex, and inflexible solutions. A software agent is an important abstraction for building distributed systems. In this paper, we provide a framework for how to construct a cooperative paradigm for network applications by intelligent communication among agents. A case study--Urban Traffic Control System--is given to show how to construct a cooperative paradigm for a distributed working environment step-by-step from bottom to top. From our experience, we found that with the solution to construct a cooperative paradigm for the agents representing different entities in the real world, a network can become a natural, efficient, and easy working environment.