The paper introduces optimal coarse correlated equilibria for continuous-time mean field games. It presents a linear programming formulation, proves existence, and develops a no-regret learning algorithm with explicit convergence rates, supported by numerical examples.