We propose Conformal Mixed-Integer Constraint Learning, a novel framework that provides probabilistic feasibility guarantees for data-driven constraints in optimization problems.