In this paper, a new method for combinatorial optimization problems by means of multi-agent system is proposed. In our method, a lot of agents explore in the solution space, locally and globally, for finding an optimal solution. A part of experimental results is also shown in order to demonstrate the effectiveness of our method.