The massive black hole binary (MBHB) path from formation on galactic scales down to their coalescence is extremely complex, and still poorly constrained, due to the huge dynamical range involved and the variety of physical processes at play on different scales. With the approval of the LISA mission, we are entering now a new era for MBHB astronomy, that requires a detailed theoretical understanding of the binary evolution. To date, studies have typically focused on different stages separately, often assuming idealized conditions and including different physical processes. I will first review the recent efforts made by the community to constrain the expected parameter space, of coalescing MBHBs, and discuss the typical limitations of current models. Then, I will present a recent ongoing effort aimed at following in a consistent way the binary inspiral from Newtonian scales down to the coalescence, which will be achieved through a general relativistic (magneto-)hydrodynamic extension of the code GIZMO. In particular, I will show the code capabilities and accuracy in modelling both the binary orbital evolution and the gas evolution, that will allow us to finally shed light on the pre- and post-merger properties of MBHBs before the advent of LISA.