XMIT, IBM's NETDATA transmit format, has been the backbone of mainframe dataset sharing and the CBT Tape community for decades, yet its binary structure remains poorly documented and almost entirely locked behind proprietary tooling. In this talk we'll walk through the open-source Python xmi library, which can parse XMI, AWS, and HET files, browse partitioned and sequential datasets, and extract members without touching z/OS. We'll then introduce our new addition: the ability to generate valid XMIT files from scratch in pure Python, opening the door to automated dataset packaging and XMIT creation baked directly into CI/CD pipelines. Along the way you'll get a ground-up tour of the NETDATA and IEBCOPY record structures so you leave understanding exactly what's inside every byte of those .xmi files you've been passing around for years.