Abstract: This paper describes the architecture of a personalized intelligent information retrieval system for Internet. The algorithms, which complement the system, are key problems in the intelligent information retrieval domain. The system is implemented using the distributed software agent techniques, the relevant feedback approach, and the multilevel approach to intelligent information filtering based on the personalized models of multi user. Also used are the online collaboration among the multi models and the retrieval methods combined customized searching with browsing query. Compared with the systems, mentioned in the references, this system has the features of adjusting to shifting user’s interests and changing source adaptively, as well as in intelligently searching good results. The results show that this system generally can be used in the electronic business and information retrieval in Internet and have high significance in theories and applications.