There has been much research about frameworks and tools to build multi-agent systems in different domains in recent years. These systems have particular features such as autonomy, distribution, sociability, cooperation and others implemented in another software entity, known as an agent. In order to achieve some previously defined goals, the agents interact between themselves to complete their tasks. One issue that arises from this kind of software is how can we ensure their dependability, considering the reliability of critical applications and the availability of those agents that play important roles with their responsibilities; i.e., how to dynamically and automatically identify the most critical agents and increase their availability and reliability? To this end, over the past few years there has been work on this problem proposing different approaches, each one solving a restricted problem involving dependability and leaving the global problem to be solved afterwards. This paper describes a solution to increase the availability of such systems through a technique of fault tolerance known as agent replication, and to increase its reliability through a mechanism of agent interaction regulation called law enforcement mechanism. The main contribution of this work is to improve the capability of calculating how critical an agent is to the system through its interactions with other agents and to provide a framework that uses this information to ensure availability and reliability.