It is, for example R package "ggbiplot" can draw ellipse around every specified group, but these ellipses are generated via multivariate normal distribution with parameters estimated from your data, so it could not reflect your data if they are non-normally distributed. I am sometimes using these just for visual differentiation of position of groups and describe this "usage" in text.
Please have a look at Rousseuw & Tukey's work on bagplots in the American Statistician.1999. This work is primarily connected with discriminant analysis but is nonparametric.