An adaptive approach for coupling detailed kinetics with computational fluid dynamics (CFD) has been formulated to permit reactive flow simulations with improved computational efficiency. The approach makes use of a set of reduced mechanisms, each of which has been developed to reproduce the mixture kinetics under well-characterized thermo-chemical conditions. In the present implementation, the KIVA-3V code is used as the CFD framework, and CHEMKIN is used to formulate the detailed kinetics that arise in the source terms of the energy and species transport equations. An efficient two-stage mechanism reduction methodology was developed that combines elemental flux analysis and the Mixed Integer Nonlinear Programming method. A highdimensional, nearest-neighbor search (NNS) method was implemented in the adaptive chemistry code to determine the appropriate reduced mechanism to use under a given set of thermo-chemical conditions. As a preliminary test of the adaptive chemistry concept, three reduced versions of a 385-species, 1895-reaction pentane oxidation mechanism were developed to reproduce the mixture kinetics in three contiguous temperature regimes. By using temperature alone as the query criterion, the adaptive approach shows great potential to improve computational efficiency while retaining acceptable accuracy. To test the more generally applicable NNS method, a total of 18 distinct reduced versions of the GRI 3.0 mechanism were generated that capture the essential methane oxidation chemistry over a wide range of equivalence ratio and temperature. The adaptive approach was evaluated by modeling compression ignition of a stoichiometric methane/air mixture in an HCCI engine. The accuracy and computational efficiency benefits of this approach have been assessed through comparisons to baseline cases involving the use of the full mechanism. Additional computational savings are realized by redimensioning the species array prior to integration to include only active species.