Inversion techniques applied to the radiative transfer equation for polarized light are capable of inferring the physical parameters in the solar atmosphere (temperature \(T\), magnetic field \({\bf B}\), and line-of-sight velocity \(v_{\rm los}\)) from observations of the Stokes vector (i.e., spectropolarimetric observations) in spectral lines. Inferences are usually performed in the \((x,y,\tau_c)\) domain, where \(\tau_c\) refers to the optical-depth scale. Generally, their determination in the \((x,y,z)\) volume is not possible due to the lack of a reliable estimation of the gas pressure, particularly in regions of the solar surface harboring strong magnetic fields. We aim to develop a new inversion code capable of reliably inferring the physical parameters in the \((x,y,z)\) domain. We combine, in a self-consistent way, an inverse solver for the radiative transfer equation (Firtez-DZ) with a solver for the magneto-hydrostatic (MHS) equilibrium, which derives realistic values of the gas pressure by taking the magnetic pressure and tension into account. We test the correct behavior of the newly developed code with spectropolarimetric observations of two sunspots recorded with the spectropolarimeter (SP) instrument on board the Hinode spacecraft, and we show how the physical parameters are inferred in the \((x,y,z)\) domain, with the Wilson depression of the sunspots arising as a natural consequence of the force balance. In particular, our approach significantly improves upon previous determinations that were based on semiempirical models. Our results open the door for the possibility of calculating reliable electric currents in three dimensions, \({\bf j}(x,y,z)\), in the solar photosphere. Further consistency checks would include a comparison with other methods that have recently been proposed and which achieve similar goals.