The KPZ (Kardar-Parisi-Zhang) universality class is a collection of two-dimensional random metric/polymer models and one-dimensional growth models that exhibit the same universal behaviour under rescaling. The directed landscape is the scaling limit of random metrics in this class. The KPZ fixed point is the scaling limit for random growth models in this class, and arises as a marginal of the directed landscape. In this talk, we will give a characterization of the directed landscape from its KPZ fixed point marginals. For a large range of models, this reduces the problem of proving convergence to the directed landscape to proving convergence to the KPZ fixed point. Joint work with Lingfu Zhang.