Fault tolerance and scalability are two fundamental issue of discovery information protocols for wide area distributed systems. In a our previous work we have introduced a discovery protocol focused on these two requirements but, although the led tests have shown the validity of adopted solution, the protocol performance are bound by the slow update mechanism of service cache, one of fundamental component of discovery algorithm. In particular, a slow update forces a long start-up time. In this paper we proposed a different update cache technique, based on mobile agents technology, no more focused on analysis of message forwarded through the network (passive update), but on autonomous research of information held in the other peers (active update). This strategy allows to rise in a fast manner the informative content of service cache improving the discovery algorithm performance.