Trankit is a toolkit for multilingual NLP recently released by the University of Oregon NLP group. It leverages XLM-Roberta, a SOTA multilingual pre-trained language model for sentence segmentation, tokenization, multi-word token expansion, part-of-speech tagging, morphological tagging, dependency parsing, and named entity recognition. To improve the memory efficiency of this architecture, Trankit also uses Adapters, which encapsulate task-specific information. Model developers can use a trainable pipeline for NLP tasks in over 100 languages or download pre-trained pipelines for 56 languages.