This chapter deals with nonlinear model recovery anti-windup (MRAW) synthesis. In considering nonlinear synthesis techniques for the feedback signal, the user can expect a tradeoff between anti-windup performance, which has the potential to improve since less structure is imposed on the anti-windup augmentation, and design and implementation complexity. The tradeoff can be extreme in the case of global anti-windup for plants that are marginally unstable. For exponentially unstable plants, the chapter describes a design that requires extra state measurements from the plant and then yields improved performance compared to that achieved using linear designs. After discussing anti-windup augmentation, the chapter examines the switching and scheduling of linear controllers, along with model predictive control for anti-windup design. It also evaluates global designs for non-exponentially unstable plants as well as designs for exponentially unstable plants that maximize the basin of attraction.