Human-centric AI wants to be explainable, robust, verifiable and ethically and morally acceptable. But how can this be achieved? The MUHAI community seeks to combine knowledge-based and data-driven AI by developing technology that allows AI systems to understand. Understanding is seen as the process of constructing a rich model from fragmented inputs (text, image, audio, action) using a variety of knowledge sources (language processing, ontologies, knowledge graphs, mental simulation, inference, perception and pattern recognition). The MUHAI repository contains documents, data and code resources for making understanding by AI systems possible and thus help to make AI more human-centric.