I think Robert has the right idea, or look in the User's Guide. I just did and didn't see anything that looks like a test for multivariate normality. Another good place to look is the Mplus Discussion forum.
Something to keep in mind, though, is if multivariate normality is an assumption you need to make. Mplus makes use of conditional normality assumptions, where normality is assumed conditional on the covariates. This may only apply to the case when all of the dependent variables are not continuous normal.