This paper addresses the distributed \( H_{\infty } \) consensus problem for an undirected network of first-order and second-order multi-agent systems, considering both external disturbances and nonconvex control input constraints. To handle the nonlinearities induced by the nonconvex constraints, a scaling factor is introduced in the analysis. This enables the transformation of the original system into an equivalent reduced-order model. Based on this reformulation, sufficient conditions are derived to guarantee consensus among all agents while ensuring the prescribed \( H_{\infty } \) performance and adherence to the respective nonconvex input constraints.