In the last years, the development of kites as a new approach to use the wind energy to produce electricity or to pull boats in order to save fuel has received attention from both the industry and academia. In this study, we propose a nonlinear model predictive control (NMPC) approach to control a towing kite based on an economic objective function in contrast to the tracking of a predefined trajectory, as it is usually done. We use a simple model of a kite that has been recently developed. The use of a simple model usually comes at a price of higher model inaccuracy. To cope with the uncertainties in the model as well as with external disturbances we propose the use of multi-stage nonlinear model predictive control. Simulation results show that the use of the multi-stage NMPC approach avoids the violation of constraints and achieves a better performance than standard NMPC under the presence of strong uncertainties.