<p>In this paper, a new distributed neuro-sliding control method is presented to solve the leader–follower tracking problem for a large class of high-order, unknown, heterogeneous nonlinear multi-agent systems with a general directed communication graph. These systems are characterized by their sufficiently general forms, the presence of completely unknown nonlinear functions that affect the dynamics of all states, and the presence of unknown external disturbances. Furthermore, the dynamics of the leader node may differ from those of the follower nodes. Moreover, we avoid typical restrictive assumptions like Lipschitz continuity or bounded nonlinearities. First, a sliding surface is investigated based on the total consensus errors of all nodes. Second, the unknown nonlinear dynamics of each node are estimated by employing adaptive neural network rules. The study develops a distributed controller that combines adaptive neural network approximations and the sliding mode technique to guarantee that all agents synchronize to the leader’s states with a bounded synchronization error in the presence of unknown disturbances. To demonstrate the stability properties, a neuro-sliding Lyapunov proof is provided. It is shown that the stability analysis is conducted by adjusting the control parameters and the coefficients present in the adaptive neural network rules. These coefficients are represented in a diagonal matrix, which simplifies the calculations required for the stability analysis. Various simulations are presented to validate the desirable performance of the proposed method in different situations.</p>