Abstract. In recent years, the production of 3D geospatial data using formats such as IFC, CityGML and GeoJSON, has increased. Visualizing this data on the web requires solving a variety of problems, such as the massive amount of 3D objects to be visualized at the same time and the creation of geometry suitable for a 3D viewer. Cesium and OGC introduced the 3D Tiles format in 2015 to solve these issues. They have created a specific format optimized for streaming and rendering 3D geospatial content, based on the glTF format developed by Khronos. The recency of the 3D Tiles format implies the need to experiment around this format and to test its interoperability with other geospatial and urban data formats. There is also the will to innovate on the organization of 3D objects in order to offer a better control on the visualization. Therefore, there is a need for an open source tool capable of converting 3D geospatial data into 3D Tiles to visualize them on the web, but also to test and develop new methods of spatial clustering and creating Levels of Detail (LoD) of urban objects. We propose Py3DTilers in this paper, an open source tool to convert and manipulate 3D Tiles from the most common 3D geospatial data models: CityGML, IFC, OBJ, and GeoJSON. With this tool, we ensure that the generated 3D Tiles respect the specification described by the OGC, in order to be used in various viewers. We provide a generic solution for spatially organizing objects and for creating LoDs, while allowing the community to customize these methods to go further in finding efficient solutions for visualizing geospatial objects on the web.