Dependability is a key issue in the deployment of every multi-agent system (MAS). Only if its services are perceived as dependable (e.g. available, reliable, secure and safe) will the MAS be considered useful. To ensure the domain specific dependability requirements, it is essential to enable the MAS to detect and react to critical states of agents. This can be done by either enabling the agent to deal with unexpected situations or by adding a fault-management component to the platform. This paper presents an event-based fault-management system and presents the results of its evaluation.