Accurate prediction of microstructured materials at finite deformation requires constitutive descriptions that retain nanoscale fidelity while remaining efficient at the component scale. We propose a two scale computational homogenization framework that couples a microscale finite element model with a nanoscale representative volume element solved by molecular statics. The scale transition satisfies the Hill-Mandel condition of macrohomogeneity using energetically conjugate stress and strain measures. The microscale boundary value problem is formulated in a total Lagrangian setting based on the Green Lagrange strain and the second Piola Kirchhoff stress and solved in static-implicit simulations by the Newton-Raphson method. At the nanoscale, volume averaged Cauchy stresses are obtained from molecular statics simulations in LAMMPS and transformed to the reference configuration through objective rotation and pull back operations. A consistent microscale tangent is evaluated by symmetric perturbations of the right stretch tensor from polar decomposition. Stress and the tangent moduli enter the material and geometric parts of the micro finite element stiffness matrix. Since the methodology is independent of the chosen interatomic potential, arbitrary material systems can be analyzed in the upscaling from nano to micro. We demonstrate that the solution algorithm for face-centered cubic metals modeled using the Embedded Atom Method (EAM), remains robust even in numerically demanding scenarios, such as defect nucleation in pristine single crystals accompanied by an abrupt and pronounced loss of stiffness.