Abstract: Multi-living agent theory has been a powerful tool for complex information system, where system living degree is an important index. With the boom of "Internet +", the tasks achieved by complex information system always possess the characteristics as high complexity, multi-stage feature variation, etc. To adapt to these features, a multi-living agent system oriented quantitative and planning method was proposed for complex multi-stage tasks, in which a system service model was built, service living degree was calculated, and a service strategy was designed based on stochastic service system theory. Results show that, the proposed method can solve the living degree quantization problem for multi-living agent system in the condition of weak prior information, validating its usefulness in practical situation.