Faced with processing time uncertainties, equipment reliability/availability, demand changes, manpower fluctuation and so on in a dynamic manufacturing environment, an organization structure for agile production scheduling and control based on MAS is put forward to achieve effective and efficient production. A hybrid multi-layer agent structure is put forward to facilitate constructing various agents. Bidding mechanism is adopted to decompose and assign manufacturing tasks that determine the cooperative relationship among agents at different levels. The integration of task assignment and scheduling is realized through capability matching and dynamic scheduling. Agent granularity and MAS model are determined dynamically according to production demands. Consequently, all resources in a manufacturing system are reorganized into an agile manufacturing network of agents to collaboratively plan, control and manage distributed production by exploiting the autonomous and cooperative characteristics of various agents.