Detailed simulation of reactive flow systems using complex kinetic mechanisms consisting of hundreds of species is a computationally demanding task. Hence the need for representing the complex chemical reactions by simple reduced models, which can retain considerable accuracy while rendering computational feasibility. Realistically, under different conditions and at different points in time, different reactions become important, which has been exploited to develop an adaptive mechanism reduction scheme such that the reduced reaction model adapts itself to the changing reactor conditions. A methodology is developed in this work to automatically construct reduced mechanisms by utilizing mathematical programming techniques. The reduced kinetic mechanisms are then analyzed for the range of conditions over which they retain their predictive capacity. These reduced mechanisms are then coupled with the reactive flow algorithm, by selecting the appropriate mechanism depending on reactor condition and integrating the corresponding reduced set of ODEs for the specified valid range. These ideas are demonstrated using the system of methane combustion in air.