Fifty years of technological advancements have witness drastic changes in the way probabilistic models are run to estimate where natural hazards may likely occur in the future. One element have remained the same through time though. Researchers and professionals in the field 1) collect and download data from a number of open repositories, 2) then load the same in a GIS environment where they pre-process the information and 3) export it in tabular form for it to be 4) externally loaded in a computing environment where data-driven models are built, whose results are exported ones more to be 5) loaded back in a GIS to visualize them in a map. Unfortunately, today this is still the most common practice, which in turn implies a lengthy process involving a number of input/output operations that could be removed if a single environment could manage most of them. With this specific reason in mind, we have developed our STGEE tool. Via STGEE, one can simply upload a shapefile of the geographic objects upon which one needs to calculate the probability of natural hazard occurrence (already internally labeled though with a presence/absence field) and then the rest is taken care by the STGEE, relying on the breadth of environmental properties available in Google Earth Engine as well as the machine learning tools already implemented there. STGEE, then loads the data, find a series of predictors already hosted in the cloud, trains (in this current version) a Random Forest classifier to estimate the susceptibility. A number of interactive visualization tools are there to explore the goodness-of-fit as well as the predictive power of the model at hand. The latter can be evaluated via spatial cross-validation routines whose spatial structure can be arbitrarily chosen, modifying a few parameters in the source code.