The early stages of relativistic heavy-ion collisions are dominated by strong classical color fields known as the Glasma, which is the precursor state to the quark-gluon plasma. At sufficiently high energies, the Glasma fields become approximately boost invariant and can be effectively described as a 2+1 dimensional system. However, this approximation is only strictly valid at mid-rapidity and typically neglects the longitudinal color structure of the colliding nuclei. In this talk I present our new approach to the initial state of heavy-ion collisions by extending the weak field approximation beyond the boost invariant limit. Our analytical calculation yields surprisingly simple results for the color fields and the field strength tensor of the 3+1D Glasma produced in the collision. We demonstrate quantitative agreement with non-perturbative lattice simulations and further investigate how longitudinal color charge correlations within the nuclei determine the rapidity profile of the Glasma. Finally, I discuss our ideas for new 3+1D initial state models using our framework.