We propose a mean field theory to account for the experimentally determined temperature dependence of protein stabilization that emerges in solutions crowded by preferentially excluded cosolutes. Based on regular solution theory and employing the Flory-Huggins approximation, our model describes cosolutes in terms of their size, and two temperature-dependent microscopic parameters that correspond to macromolecule-cosolute and bulk solution interactions. The theory not only predicts a "depletion force" that can account for the experimentally observed stabilization of protein folding or association in the presence of excluded cosolutes but also predicts the full range of associated entropic and enthalpic components. Remarkably, depending on cosolute identity and in accordance with experiments, the theory describes entropically as well as enthalpically dominated depletion forces, even those disfavored by entropy. This emerging depletion attraction cannot be simply linked to molecular volumes. Instead, the relevant parameter is an effective volume that represents an interplay between solvent, cosolute, and macromolecular interactions. We demonstrate that the apparent depletion free energy is often accompanied by significant yet compensating entropy and enthalpy terms that, although having a net zero contribution to stabilization, can obscure the underlying molecular mechanism. This study underscores the importance of including often-neglected free energy terms that correspond to solvent-cosolute and cosolute-macromolecule interactions, which for most typical cosolutes are expected to be temperature dependent. We propose that experiments specifically aimed at resolving the temperature-dependence of cosolute exclusion from macromolecular surfaces should help reveal the full range of the underlying molecular mechanisms of the depletion force.