Abstract: It is proposed that mobile agents act as carriers of tasks of applications or services in dynamic heterogeneous grid environments because of autonomy and intelligence of mobile agents. And a new TSP model—dynamic market TSP model, is introduced that is more suited for grid and a path optimization algorithm is proposed to optimize the migration paths of mobile agents in this model. The simulation experiments in different TSP maps have been conducted for this algorithm. The data show that this algorathm makes mobile agents run through less distance for their tasks than the traditional optimization algorithm does, and it converges more quickly.