We develop a theoretical framework to elucidate the mechanistic basis of thermal niche partitioning in ectotherms. Using a food web module of two consumers competing for a biotic resource, we investigate how temperature effects on species' attack and mortality rates scale up to population-level outcomes of coexistence and exclusion. We find that species' differences in competitive effects arise from asymmetries generated by the non-linear temperature response of mortality: cold-adapted species experience stronger intra-specific competition than warm-adapted species; they also exert weaker competition on, and experience stronger competition from, warm-adapted species. These asymmetries become greater as seasonal fluctuations increase, generating latitudinal variation in coexistence and priority effects. Characterizing species' thermal niches in terms of mechanistic descriptions of trait responses allows for testable predictions about population-level competitive outcomes based solely on three fundamental, and easily measurable, quantities: attack rate optima, response breadths and temperature sensitivity of mortality. We test our predictions with data from an insect host-parasitoid community. By quantifying the three basic quantities we predict that priority effects cannot occur, which is borne out by population-level experiments showing that the outcome of competition does not depend on initial conditions. More broadly, our framework can predict the conditions under which exotic invasive species can exclude, or coexist with, native biota, and the effects of climate warming on competitive communities across latitudinal gradients.