We present first synthetic tests for a newly developed 3D joint inversion framework for seismic refraction, magnetotelluric (MT) and gravity data. Using this framework we can investigate different joint inversion approaches and adjust the parametrization and coupling between the different methods to the individual targets. Tests with a simple synthetic model show that even a small amount of additional MT and gravity data can provide substantial improvement over a single tomographic inversion. As expected, utilizing the correct parameter relationship in the joint inversion gives better results than a purely structural coupling. Currently we are constructing a synthetic salt dome model to test the joint inversion algorithm on realistic geometries and we will apply our approach to a dataset from the Nordkapp Basin.