We study a Monte Carlo algorithm for computing marginal and stationary densities of stochastic models with the Markov property, establishing global asymptotic normality and () convergence. Asymptotic normality is used to derive error bounds in terms of the distribution of the norm deviation.