Increasingly, multi-agent systems are being designed for a variety of complex, dynamic and uncertain domains. An agent in such domains often needs to learn or evolve to adjust its behaviors, or negotiate with other agents, and so on. Immune cells are the key components of the human immune system, their evolutions in the whole lifecycle directly decide the immune system?s performance. In this paper, a model of immune evolution multi-agent system (IEMAS) is introduced based on immune cells? evolution principles. In such a system agents can evolve to adapt to the complex, dynamic and uncertain environment. Also, the formal model of IEMAS is described, and then the mature immune cell evolution algorithm and the memory immune cell evolution algorithm of IEMAS are presented. Finally, the run of IEMAS applied to the task allocation in multi-agent systems are presented and displays the efficiency of IEMAS in the complex, dynamic and uncertain environment.