For many reasons, predictable coordination of autonomous computing entities under time constraints is a demanding research subject. Devising a general framework that is able to cope with the somewhat contradictory requirements such as timeliness and adaptability is a challenge. This paper presents the preliminary results of an ongoing research project dealing with the application of multi-agent systems in real-time applications. The paper focuses on a well known coordination protocol, the contract network protocol, and proposes some extensions to it in order to tackle real-time systems. The developed concepts are illustrated using a case study of a robotic system.