Protein folding, a process that spans a wide range of timescales and that involves complex conformational dynamics, is an extremely challenging problem to decode at the atomic level. Over the past decade, coarse-grained (CG) models that rely on a reduced representation of the polymer chain as dictated by the native structure have been quite successful in characterizing and predicting a variety of aspects of the folding mechanism of single-domain proteins. The ever-increasing ability of this minimalist treatment is a primarily due to the rapid and efficient sampling afforded by coarse-graining that smoothens the folding landscape and the simple nature of the constituent physical energy functions that can be easily cast in various forms or parameterized using experimental or knowledge-based approaches. With the advances in computational power we have now reached a stage where CG simulations can be routinely performed to test various mechanistic hypothesis, to interpret intricate experimental observables and even suggest new experimental avenues. Here, we provide an overview of recent CG developments that have predicted experiments quantitatively and others that have sought to use experimental information as constraints to tune the energetics and answer fundamental questions in folding and conformational behavior of proteins. We further discuss open issues and point to new directions that can drive the CG models toward better agreement with experiments and to a better understanding of folding mechanisms in general. © 2012 John Wiley & Sons, Ltd.