Data mining applications today are more likely to deal with distributed data. One of the challenges is to protect the privacy of local data from being exposed to other sites. Various approaches have been reported in the literature, but we have found no work using the mobile agent approach to tackle this problem while mobile agents are considered very suitable for distributed computing tasks. In this paper, we propose an agent-based approach to mine association rules from data sets that are distributed across multiple locations while preserving the privacy of local data. This approach relies on the local systems to find frequent itemsets that are encrypted and the partial results are carried from site to site. We present a structural model that includes several types of mobile agents with specific functionalities and communication scheme to accomplish the task.