Two-dimensional transition metal carbides (MXenes) have been demonstrated to be promising supports for single-atom catalysts (SACs) to enable efficient and effective oxygen evolution reaction (OER). However, the rational design of MXene-based SACs depends on an experimental trial-and-error approach. A theoretical guidance principle is highly expected for the efficient evaluation of MXene-based SACs. Herein, high-throughput screening was performed through first-principles calculations and machine learning techniques. Ti3C2(OH)x, V3C2(OH)x, Zr3C2(OH)x, Nb3C2(OH)x, Hf3C2(OH)x, Ta3C2(OH)x, and W3C2(OH)x were screened out based on their excellent stability. Zn, Pd, Ag, Cd, Au, and Hg were proposed to be promising single atoms anchored in MXenes based on cohesive energy analysis. Hf3C2(OH)x with a Pd single atom delivers a theoretical overpotential of 81 mV. Both moderate electron-deficient state and high covalency of metal–carbon bonds were critical features for the high OER reactivity. This principle is expected to be a promising approach to the rational design of OER catalysts for metal-air batteries, fuel cells, and other OER-based energy storage devices.