Despite the numerous possible applications of wireless sensor networks (WSNs), there is a key disadvantage related to the high complexity in programming WSNs, which is a result of their distributed and built-in features. To overcome this shortcoming, software agents have been identified as a suitable programming paradigm. The agent-based approach commonly uses a middleware for the execution of the software agents. In this regard, the present paper aims at comparing Java-based agent middleware platforms in their performance for the WSN application domain. Experiments were performed to analyze two versions of tracking applications, based on different agent models implemented for a given set of middleware platforms that support programming at a high-level of abstraction. The results highlight the differences in the resource consumption (CPU, memory, and energy) and in the communication overhead, providing an indication of suitability for each type of analyzed middleware, considering specific concerns while developing WSN applications. © 2001-2012 IEEE.