We present modosc, a set of Max abstractions designed for computing motion descriptors from raw motion capture data in real time. The library contains methods for extracting descriptors useful for expressive movement analysis and sonic interaction design. modosc is designed to address the data handling and synchronization issues that often arise when working with complex marker sets. This is achieved by adopting a multiparadigm approach facilitated by odot and Open Sound Control to overcome some of the limitations of conventional Max programming, and structure incoming and outgoing data streams in a meaningful and easily accessible manner. After describing the contents of the library and how data streams are structured and processed, we report on a sonic interaction design use case involving motion feature extraction and machine learning.