With the recent increase demand for cloud services, handling clients’ needs is getting increasinglychallenging. Responding to all requesting clients with no exception could lead to security breaches, andsince it is the provider’s responsibility to secure not only the offered cloud services but also the data, it isimportant to ensure clients reliability. Although filtering clients in the cloud is not so common, it isrequired to assure cloud safety.In this paper, we made use of multi agent system interaction mechanisms to handle the cloud interactionsfor the providers’ side, and of Particle Swarm Optimization to select the and determine the accurate trustweight and therefore to filtrate the clients by determining malicious and untrustworthy clients. Theselection depends on previous knowledge and overall rating of trusted peers. The conducted experimentsproves that the combination between MAS, PSO and trust outputs relevant results with different number ofpeers, the solution is able to converge to the best result after small number of iterations. To explore morespace and scenarios, synthetic data was generated to improve the model’s precision .