This paper deals with a multi-agent technology as a tool to introduce an autonomic behavior in network control and management. In a first section, we describe the autonomic communication paradigm, the characteristics, and the need for new architectures. Then, we focus on the differences between a classical and an autonomic network. Some related works involving the combination of artificial intelligence and telecommunication networks are presented. Then, characteristics of agents and multi-agent systems are presented to provide the network with autonomic behavior. A specific section also analyzes the different proposals in the literature. Then, a network-oriented platform associated with agent capabilities is described. The following section presents an application of the architecture toOspf weight setting optimization. Finally, the paper details two experiments on theOspf weight setting applied to a small network and to a large scale network.