Recently, collaborative filtering is proposed as an information gathering technology of the WWW in the network. Collaborative filtering is a technology that recommends information on the Web page for an arbitrary user who wants to acquire recommendation information based on many users? evaluation and retrieval histories. In this paper, the bookmark shared system that filters bookmark information collaboratively is described. Especially, we explain the details of the bookmark shared system using agent systems and it?s collaborative filtering on the distributed environment. In a concrete agent modeling, the multiagent does a simple communication to notify the profile update, and the retrieval processing is done by a mobile agent. As a result, the profile management on the distributed environment is facilitated, and it is possible to construct collaborative filtering system that can decrease the communication frequency in the retrieval processing.