Open die forging is a manufacturing process with a number of advantages; in particular it is an inherently flexible manufacturing process that makes efficient use of raw material. A fundamental drawback of this process, however, is the difficulty found in creating forging programs to control part manipulation and forming steps. A-priori approaches to creating these programs, such as by using FEM simulations or using modeling materials, are slow and have a strong tendency for errors to accumulate when predicting the results of consecutive forming steps. In this paper we present a new approach in which process feedback is used between forming steps to update a part geometry model that allows the forming sequence to be adjusted adaptively. This approach has been implemented in a simulated forging cell that uses non-linear FEM analyses to predict the effects of each forming step. A fully adaptive control scheme has been implemented that efficiently forges bars of one cross sectional shape into another shape, such as square to round or hexagonal. Programming the forging system with this scheme has proved particularly simple; the shape of the raw material is measured, and a desired shape is specified. Physical experiments have confirmed the simulation results.