The augmentation of Cyber-physical Systems (CPSs) with Digital Twins (DTs) through various twinning architectures has significantly changed how complex systems are designed, operated, and maintained. Recent advances in twinning provide high-fidelity digital replicas of physical entities, facilitating enhanced monitoring and predictive services. However, creating effective twinning systems for distributed CPSs is difficult due to complex relationships and convoluted interactions between system components. This makes monitoring and ensuring the system adheres to the requirements more challenging. Machine learning-based techniques have been proposed, but these have scalability and performance limitations. This work proposes a multi-agent-oriented architecture aimed at overcoming these challenges by fostering the integration between CPSs and their DTs. Our architecture enhances system monitoring and anomaly detection by combining intelligent software agents with a complex event-processing engine. This ensures robustness and fault tolerance in CPS applications while improving operational efficiency and accuracy. Besides describing our proposed framework, this paper discusses its implications by means of a multi-robot system for a warehouse, which demonstrates the applicability of our solution and provides some insights into future directions for research in this domain.