Abstract: An interactive system for E-retail is designed to meet the requirements of fashion marketing development. The infrastructure and prototype implementation of this system are presented and analyzed. The interactive system is composed of three fundamental modules: recommendation, negotiation and feedback, and is applicable for fashion E-retail. Agent technique is applied on the process of interaction for rapid respondence in this system. To provide the personalized internet service, the system can not only generate real-time recommendation according to the characters of customers, but also collect feedback information and make the dynamic adaptation of customer profiles based on behavioral data and domain knowledge models. This approach is domain-independent and applicable for a wide-variety personalized market of fashion E-retail.