In the field of mobile cloud computing, trust security in mobile cloud computing is very important because data is stored and processed in remote servers. Facing the security and trust problems of mobile agent system in cloud computing environments, we use human trust mechanism(HTM) to study the formation, propagation and evolution of subjective trust. Based on this, we propose the mobile agent subjective trust management algorithm(MASTM). We compute the mobile agent's interaction experience with the host as well as the underlying trust data collected by third parties. Moreover, we construct a trusted host selection algorithm, an isolated malicious host algorithm and a comprehensive trust degree calculation algorithm based on these data. These algorithms implement the function of selecting a trusted fleet of machines and isolating malicious hosts, enhancing the security interaction between mobile agents and hosts. Moreover, we conduct simulation experiments for all the given algorithms to verify the feasibility and effectiveness of the algorithms.