We generalized the Euler deconvolution method to a joint scheme, which consists of locating the horizontal and vertical positions of the top of potential-field 3D sources. These results were then used to constrain the depth to the top of the models obtained by cross-gradient joint 3D inversions, imposing fixed known values in the a priori models. The coupling of both methods produced more realistic density and magnetization models for separate and joint inversions, relative to those obtained by applying cross-gradient joint inversion only. This strategy was tested on a 3D synthetic experiment, and on a real field data set from the northwest region of the Baja California Peninsula, Mexico. After locating the vertical position of the source, the algorithm uses this information to obtain density and magnetization models that enhanced their structural compatibility and reduces the ambiguity on the interpretation of their structural characteristics laterally and at surface.