We propose a structure-based protocol for the development of customized covalent inhibitors. Starting from a known inhibitor, in the first and second steps appropriate substituents of the warhead are selected on the basis of quantum mechanical (QM) computations and hybrid approaches combining QM with molecular mechanics (QM/MM). In the third step the recognition unit is optimized using docking approaches for the noncovalent complex. These predictions are finally verified by QM/MM or molecular dynamic simulations. The applicability of our approach is successfully demonstrated by the design of reversible covalent vinylsulfone-based inhibitors for rhodesain. The examples show that our approach is sufficiently accurate to identify compounds with the desired properties but also to exclude non promising ones.