Having standardized metadata is the first step toward deploying smart building applications over heterogeneous buildings. Such a conversion process is highly manual because of different conventions in existing building metadata and diverse building configurations. Many machine learning methods have been attempted to ease the process by reducing the amount of experts’ training examples and reusing the knowledge in different data sets. However, many of the end-users, such as building managers and commissioning practitioners, are unfamiliar with machine learning and programming interfaces. We implement and demonstrate a web-based graphical user interface whose workflow is designed based on a common programming interface, Plaster, for building metadata normalization. We implement three algorithms, Zodiac, BuildingAdapter, and Scrabble, though any new algorithms can be added. Users are instructed for proper actions with information visualization at each step to easily complete the procedure. The service is freely available at https://plaster.ucsd.edu.