Nonparametric regression models involving circular variables (as responses and/or covariates) are revised in this chapter, focusing on a kernel approach. For linear response and circular covariate, both suitably adapted local polynomial and local likelihood models are introduced. The local polynomial approach is also presented for circular response regression. The use of these estimators for testing different hypotheses is also revised and their applicability is illustrated with a real data example. Finally, we provide some resources for the computation of the different tools.