FitPDF fits complex mixture models of various distributions to observational data. This is useful, for instance, for characterizing the observed pulse energy distributions of radio pulsars and repeating fast radio bursts (FRBs). However, FitPDF is data-agnostic and can, in principle, fit any distribution data and determine its underlying probability density function (PDF). The FitPDF software suite includes tools for simulating distribution data, fitting complex mixture distributions to data, and comparing the resulting fits in an information-theoretical sense. The code supports several model distributions, which are homogeneous or heterogeneous mixtures of individual component distributions (e.g. normal, lognormal, power law). FitPDF, based on PyMC (ascl:1610.016), operates in a Bayesian and unbinned manner, thereby maximizing the information extracted from the data and robustly determining the model parameters and their uncertainties.