The use of mobility has been shown to be beneficial in wireless ad hoc and sensor networks, for improving communication performance and other functionality. This paper discusses the communication throughput, goodput and delay considerations when a set of mobile nodes is used as relays to transfer data among multiple static nodes. While previous work has considered randomly mobile nodes, we consider controlled mobile agents. Randomly mobile agents are not available in many network scenarios, such as embedded sensor network deployments, and the use of controllably mobile agents has been considered for such networks. We derive results for the worst case delay, throughput and goodput with controllably mobile relays. Our analysis indicates that this scenario differs fundamentally from the random mobility case. This scenario could, however, be used in defense applications for better communications yield. Further, our results are guaranteed to be achieved in a particular topology, as opposed to previous results which are probabilistic for a particular deployment. We also discuss practical algorithms that can be used to control the routes of mobile agents.