In the common case of a diagonal covariance matrix, the multivariate PDF can be obtained by simply multiplying the univariate PDF values returned by a scipy.stats.norm instance. If you need the general case, you will probably have to code this yourself (which shouldn't be hard).