Cooperative control of multiagent systems (MAS) on communication networks has received a great deal of attention, mostly for the case of homogeneous agents, which all have the same dynamics. An advantage of cooperative synchronization mechanisms is their local distributed nature, which makes them scalable to large networks. However, most existing design mechanisms require some global information, such as the leader's dynamics or global graph information, so that the control protocols are technically not fully distributed. Moreover, the distributed nature of the control protocols makes them susceptible to faults or uncertainties. In this paper, we study heterogeneous MAS, where all agents may have different dynamics. We provide adaptive resilience mechanisms for rejecting actuator faults, and guarantee exponential convergence of synchronization errors, whereas most existing results on actuator faults guarantee only boundedness of errors. Finally, we provide algorithms that are fully distributed, requiring no knowledge of either the leader's dynamics or of graph properties.