The dynamic interactions within a social network can be modeled with respect to service transactions. In this paper, we develop a framework for modeling a service agent network in an attempt to study the emergent dynamic behavior (e.g., performance and formation) of the network. In the framework, a service transaction process is introduced that is composed of such steps as service matching, service evaluation, service propagation, and agent profile evolution. Through the transaction process model, the dynamic performance and formation of the agent network can readily be simulated. Our simulation results show that a network of service agents that select their cooperative partners based on high cumulative profits can yield high service throughputs.