Multi-agent scheduling offers reactivity and distributed decision-making for floor shop control. Agents, which may represent any entities that act in production, negotiate to find best schedules. In this paper, we present a new multi-agent scheduling method that integrates both planning and maintenance activities. Actually, more than one plan can be generated for a job production. We suppose that plan selection must depend on information about machines maintenance and states to offer realistic schedules. Tests demonstrate that despite increasing time resolution with the agents’ number, our system is scalable with reduced Cmax and machines failure risk.