In this paper, the problem of qualitative description and quantitative optimization for tactical reconnaissance agents system organization is considered with objective of higher teamwork efficiency and more reasonable task balancing strategies. By analyzing tactical reconnaissance system and its environment, task-(role)-entity agent mapping mechanism and agents in system organization, the system framework is qualitatively described. By transforming the system into an interaction task request-service mechanism queuing system, a Markov chain of system state transition is obtained, since its state transition process in interaction is Markov process and accords with real tactical reconnaissance behaviors. By solving the state transition equations, the inherent relationship of tactical reconnaissance agents is found and the optimized system configuration is obtained. The established simulation demonstration system proves that the proposed approach and model are feasible and effective.