This paper addresses a fault-tolerant formation tracking problem for multiple mobile robots, where the dynamics of each robot is nonholonomic with time-varying input delays and actuator faults. The control problem is challenging since the actuator faults (e.g., loss of effectiveness and biased faults) are unknown in the time-delay multi-agent systems. To this end, an adaptive compensator is proposed for the uncertain time-delay systems induced by the loss of effectiveness faults. With the aid of neural networks and auxiliary variables, a novel adaptive fault-tolerant formation tracking controller for each robot is developed without any restrictions on the rate of their unknown variables induced by the biased faults. The uniformly ultimately boundness for the closed-loop system is derived according to the Lyapunov–Krasovskii functional. Numerical simulations verify the effectiveness of the proposed method.