Ambulatory health care centers are the major arena for delivery of primary health care to patients. In contrast to hospital or urgent care scheduling systems, ambulatory health care centers offer different challenge for scheduling optimization. The aim is similar, i.e. to reduce the average waiting time and maintain high resource utilization, but the concept is different. In this paper we present a multi-agent based system for long-term scheduling, where agents use their time-series forecasting and pattern recognition abilities to predict possible patient flow peaks and inform the main scheduling agent of these events. To ease such peaks, we use adaptive nurse scheduling by adapting the nurse schedules each week-end. Our multi-agent system helps the personnel managers and other staff in health centers to adapt the nurse and patient scheduling in dependence of the current patient flow.