Abstract: Case based reasoning (CBR) methodology is increasingly used for the development of collaborative interface agents. Almost all-existing CBR based collaborative interface agents are implemented according to the thin client model where the CBR engine is executed on a central server to whom the distant clients (agents) are connected. Such an architecture suffers from a number of well known drawbacks such as poor response time, low dependability, but also poor privacy protection features. To overcome drawbacks of the above-mentioned architecture, some distributed alternatives have been proposed in the CBR literature. However we are not aware of any work that presents a methodological comparison study of possible distributed CBR application architectures. This paper aims at filling this gap.