A peer-to-peer (P2P) system consists of a number of decentralized distributed network nodes that are capable of sharing resources without centralized supervision. Many applications such as IP-phone, contents delivery networks (CDN) and distributed computing adopt P2P technology into their base communication systems. One of the most important functions in P2P system is the location of resources, and it is generally hard to achieve due to the intrinsic nature of P2P, i.e. dynamic re-configuration of the network. We have proposed and implemented an efficient resource locating method in a pure P2P system based on a multiple agent system. The model of our system is a distributed hash table (DHT)-based P2P system that consists of nodes with DHT (high performance nodes) and nodes without DHT (regular nodes). All the resources as well as resource information are managed by cooperative multiple agents. In order to optimize the behaviors of cooperative multiple agents, we utilize the ant colony optimization (ACO) algorithm that assists mobile agents to migrate toward relatively resource-rich nodes. Quasi-optimally guided migrating multiple agents are expected to find desired resources effectively while reducing communication traffic in the network. Efficient migration is achieved through the clustering of nodes that correlates nodes into a group by logical similarity, and through an indirect communications that are typical of social insects, called stigmergy. When an agent finds a resource-rich node, it strengthens the path toward the node so that further efficiency is gained. Strengthening of the route is achieved by pheromone laid down by preceding agents that guides succeeding agents. The numerical experiments through simulation have shown a significant reduction of generated messages.