Contour tracking accuracy is the key performance index of a multi-axis motion control system. In the previous design of the contour error feedback controller, the independent design idea of the tracking controller and contour controller of each logic axis is often adopted, which is not easy to realize the optimal control of tracking accuracy and contour accuracy. The coupling increment of each controller is inevitable in motion control compensation, which impacts the control system’s stability. Here, we propose a predictive control method for contour error. At first, through the state space model of the feed system, the system dynamics are judged to predict the contour accuracy in the future finite time domain. Besides, the linear quadratic performance index function of contour error is established, and the performance of the control objective is continuously optimized during the limited time domain with the form of rolling optimization. A closed-loop solution is obtained, which is most suitable for the current motion situation and could add to the system to realize the optimal control considering the trajectory contour and the tracking stability of each servo axis in the subsequent few sampling cycles. Finally, utilizing comparative experiments on the three-axis motion control platform, it is verified that the proposed contour predictive control strategy not only effectively reduces the tracking error but also significantly improves the contour accuracy, which is suitable for application in a complex industrial environment.