Owing to the increasing complexity of traditional real-time systems, there is a growing interest in applying AI techniques, and in particular knowledge-based systems, in this domain. However, the time and memory-consuming AI algorithms, with unpredictable or highly variable performances, are usually not compatible with the strict requirements of realtime systems. The paper describes a software environment well adapted to the development and delivery of real-time knowledge-based systems. This environment is built around a multi-agent, multi-task realtime architecture based on the blackboard model. The main aspect of the developed architecture is its ability to provide guaranteed response times, through the use of various scheduling techniques, among which is progressive reasoning.