The evolution of AGB stars is notoriously complex. The confrontation of AGB population models with observed stellar populations is a useful alternative to the detailed study of individual stars in efforts to converge towards a reliable evolution theory. I review here the impact of studies of star clusters on AGB models and AGB population synthesis, deliberately leaving out any more complex stellar populations. Over the last 10 years, despite much effort, the absolute uncertainties in the predictions of the light emitted by intermediate age populations have not been reduced to a satisfactory level. Observational sample definitions, as well as the combination of the natural variance in AGB properties with small number statistics, are largely responsible for this situation. There is hope that the constraints may soon become strong enough, thanks to large unbiased surveys of star clusters, resolved colour-magnitude diagrams, and new analysis methods that can account for the stochastic nature of AGB populations in clusters.