We use a novel projection-based Lyapunov function to prove the first asymptotic optimality theorem for fully heterogeneous weakly-coupled Markov Decision Processes.