We consider a system of agents as a layer between services and the network hosts. We present a self-organizing system of agents able to offer services with an increase of the speed, and decrease of the network traffic. A coordination model defines how the agents interact, and how their interactions can be controlled; this includes dynamic creation and destruction of agents, control of communication flows among agents, control of spatial distribution of agents, as well as synchronization of actions over time. Various distances and topological relations are used to explain our self-organizing mechanism. We analyze the stability of the system, and provide a new result regarding a weak form of stability depending only on the moment of the first requests of agents by hosts. Finally we provide a visualization of the services evolution using self-organizing maps, and presenting a service clustering structure.