The main hurdle in instrumentalizing agricultural soils to sequester atmospheric carbon is the development of methods to measure soil carbon stocks which are robust, scalable, and widely applicable. Our objective is to develop an approach that can help overcome these hurdles. In this paper, we present the Wageningen Soil Carbon STOck pRotocol (SoilCASTOR). SoilCASTOR uses a novel approach fusing satellite data, direct proximal sensing-based soil measurements, and machine learning to yield soil carbon stock estimates. The method has been tested and applied in the USA on fields with agricultural land use. Results show that the estimates are precise and repeatable and that the approach could be rapidly scalable. The precision of farm C stocks is below 5% enabling detection of soil organic carbon changes desired for the 4 per 1000 initiative. The assessment can be done robustly with as few as 0.5 sample per hectare for farms varying from 20 to 150 hectares. These findings could enable the structural implementation of carbon farming.