We introduce bnRep, an open-access, machine-readable repository that collects Bayesian networks from the academic literature. Networks in bnRep span 43 different fields of application, with a total of 530 networks associated with 388 different publications from 1988 to 2024. The repository is aimed to serve as a ground truth for the development and benchmarking of structure learning algorithms, analysis of structural properties of Bayesian networks across different domains, and education. We provide a comprehensive review of the academic literature for Bayesian networks with up to 500 variables. The structure and metadata of all networks are standardized and accessible through a dedicated R package. We discuss the properties of the networks in the repository and identify some challenges for the development of novel structure learning algorithms.