Summary: This paper presents a consensus-error-tracking iterative learning control method to tackle the consensus problem for a class of leader-following nonparametric uncertain multi-agent systems, which perform a given repetitive task over a finite interval with arbitrary initial error. The iterative learning controllers are designed by applying Lyapunov synthesis. As the iteration increases, each following multi-agent’s consensus-error can track its desired consensus-error trajectory, and the all following multi-agents’ states perfectly track the leader’s state on the specified interval. The robust learning technique is applied to deal with the nonparametric uncertainties, and the hyperbolic tangent function is used to design feedback terms, in order to compensate the cycle-varying but bounded uncertainty. Numerical results demonstrate the effectiveness of the learning control scheme.