A rapid growth of available geospatial data requires development of systems capable of autonomous data retrieval, integration and validation. Mobile agent technology may provide the suitable framework for developing such systems since this technology can deal, in a natural way, with the distributed heterogeneous nature of the data. In this paper, we evaluate our novel multi-agent architecture for geospatial data integration and compare its performance with a client/server and a single-agent architecture. We analyze the performance alteration for various numbers of participating nodes, amount of database accesses, processing loads, and network loads.