Mobile agents are an extension of multiagent systems in which the agents are provided with the ability to move from node to node in a distributed system. While it has been shown that mobility can be used to provide simple, efficient, fault-tolerant solutions to a number of problems in distributed computing, mobile agents have yet to become common in mainstream applications. One of the reasons for this may be the lack of an algorithmic theory which would provide a framework in which different approaches can be analyzed and the limits of mobile agent computing explored. In this paper we attempt to provide such an algorithmic theory.