Summary: This paper investigates the distributed consensus tracking problems of multi-agent systems on undirected graph with a fixed topology. Each follower is assumed to be in strict-feedback form with unknown state-dependent controlling effects. A distributed robust adaptive neural networks-based control scheme is designed to guarantee the consensus output tracking errors between the followers and the leader are cooperatively semi-globally uniformly ultimately bounded. Command filtered backstepping technique is extended to the consensus tracking control problems, which avoids the classical “explosion of complexity” problem in standard backstepping design and removes the assumption that the first \(n\) derivatives of the leader’s output should be known. The function approximation technique using neural networks is employed to compensate for unknown functions induced from the controller design procedure. Stability analysis and parameter convergence of the proposed algorithm are conducted based on algebraic graph theory and Lyapunov theory.