improb (a python module for working with imprecise probabilities). The library supports arbitrary finitely generated conditional lower previsions, belief functions, linear-vacuous mixtures, probability measures, n-monotone lower probabilities, Mobius transforms, and Choquet integration. Various decision criteria, such as Gamma-maximin, Gamma-maximax, interval dominance, and maximality, are implemented. For sequential decision problems, the library has a convenient interface for constructing decision trees of any size, and has algorithms for solving them by normal form, or by normal form backward induction.