Next-generation real-time distributed systems present increasing demands for flexibility and adaption from computing objects, due to dynamic requirements and complexity. To meet those demands for flexibility and adaptation and at the same time ensure a predictable temporal behavior is a challenge that the great majority of existing methods can not cope with. Key properties like optimization of global utility, fairness and timeliness must be taken in account, specifically at the communication and coordination levels. In this paper, preliminary results of an ongoing research project which aims to provide a framework for developing real-time distributed applications using the concept of real-time agents are presented. The approach focuses on aspects such as agent communication and coordination scemes. It combines known results from distributed agent systems, such as the contract net protocol for selecting the most suitable agent for a solving a given problem, with real-time message and task scheduling algorithms. In order to illustrate the concepts proposed, a robotic systemis presented as a case-study.