The paper gives overview of a multi-agent real-time scheduler for the European operation of one of the largest rent-a-car company in the world. It describes requirements for scheduling of cars and drivers and outlines main features of the ontology-based, multi-agent approach, including systems architecture and performance measurements. The key design decisions and results of the first stage of the system development are also covered. The system is capable of scheduling complex interdependent operations of a large number of resources and of updating schedules affected by the occurrence of unpredictable events in real time. The multi-agent approach developed for scheduling of car rentals can be applied to a variety of complex real-time scheduling and optimization applications.