This task defines a transformation from the previously extended observation model to the actual source code that implements the monitoring infrastructure. Since the implementation is highly specific for the specific system and target implementation platform (e.g., the used multi-agent system or middleware), these rules have to be tailored for the target platform and system. However, some of the basic principles remain the same, regardless of the transformation target.

The classes and sequence diagrams have to be translated into the target programming language and the target platform, i.e., the multi-agent platform or middleware the system will run on. Sequence diagrams become implementations of the methods of the classes. The flow of information from the agents to the observers and from there to the controllers has to be captured as defined in the observation model and adapted to the communication infrastructure provided. A decision has to be made, whether or not observers and controllers are independent
agents or become part of the agents defined in the domain model, whether properties of the agents can be accessed directly or only via message passing, etc.