It is well known that the relatively high overpotentials for oxygen evolution reactions arise from similar binding energies for HO* and HOO* intermediates that tend to move in lockstep on different catalysts. A strategy to break this scaling limitation is through differential stabilization of the two intermediates by engineering the reaction environment. Nanoporous materials such as zeolites and metal-organic frameworks provide such a possibility through their well-defined pore space that presents microscopically tunable reaction environments. Yet the predictive understanding of catalytic actions in these heterogeneous catalysts requires the ability to describe not only the electronic structure at the active site, but also how the extended environment interacts with the reacting species through individually weak yet collectively significant non-covalent interactions. In this talk, we present a new computational procedure that couples quantum-chemical methods and forcefield-based sampling to enable a consistent calculation of local activation energies for articulated reactants. As a first application, we studied hydrocarbon reactions in four different zeolites and showed how the method allows for the estimation of transition-state entropy and for the systematic calculation of site-dependent activation energies. The latter can be used to derive rigorous ensemble-averaged barriers for comparison with experiments, as well as guiding future strategies for controlling Al distributions during zeolite synthesis.