Abstract: In order to implement Web service selection effectively, a service selection model based on multi-agent collaboration is constructed. Communication performatives in multi-agent system for service selection are designed, by which agents perform collaboration. The evaluation algorithm based on fuzzy matching is proposed, which provide consistent measure for evaluation criteria. The preferences of service requestors are aggregated according to the voting protocol, by which the right value of each evaluation item it taken. Experimental results demonstrate that the mean effectiveness of service selection is increased by 29 percent; the average selection time decreased by 58 percent, and the computing complexity drops from <em<O(n<sup<5</sup<)</em< to <em<O(n<sup<2</sup<)</em< in the model.