Based on designing of reverse operator integer coding, the artificial wolves’ position and intelligent behaviors are redesigned. Then combined with the probabilistic nearest neighbor method, a discrete wolf pack algorithm(DWPA) is proposed to solve traveling salesman problem(TSP). The DWPA preserves the cooperative searching feature based on job distribution of the wolf pack, and makes a good balance between the searching ability in breadth and depth. Algorithm efficiency of TSP problems against five existing algorithms is examined for C-TSP and some TSPLIB instances of various sizes. Simulation results show that the DWPA has more advantages on accurate rate, computational robustness and iteration number.