Health status management of heart failure patients is a rather complex and difficult procedure. Currently, a well-promising approach facilitating this aim is home care systems, which provide monitoring, educational and surveillance services over several communication platforms, guaranteeing continuity and quality of care. In this paper, a multi-agent system architecture is proposed to support heart failure management in a generic home care setting. The system incorporates various processing modules based on statistics, rules and models that are applied to the medical data of patients, as well as learning capabilities based on the medical interventions that take place, and negotiation schemes among agents. The ultimate goal of this system is the characterization of the patients' health status and accordingly the notification of the corresponding medical personnel to take medical actions.