This paper investigates the semi-global cooperative cluster output regulation problem of heterogeneous multi-agent systems with input saturation, the exosystems for each cluster can be different. To avoid using global information (e.g., the minimal nonzero eigenvalue of the Laplacian matrix) in the control protocol, an adaptive dynamic compensator is proposed to estimate exosystem’s state in fully distributed manner. A dynamic event-triggering mechanism with adaptive parameter is proposed in order to reduce the usage of communication resources. Low-gain feedback technique is utilized to deal with the influence of input saturation, and Lypunov-based stability analysis results are obtained. Moreover, it is formally shown that Zeno behavior can be excluded. The superiority of the proposed methods includes: the agents in each cluster are also heterogeneous, which is essentially different from [1]; the event-triggered control strategy does not depend on any global information; and the influence of saturation nonlinearity can be eliminated with low-gain feedback. Finally, a numerical example is provided to illustrate the effectiveness of the proposed methods.