We introduce dQP, a solver-agnostic, modular framework for differentiable quadratic programming (QP), enabling plug-and-play differentiation with any solver and demonstrating advantages in large-scale sparse problems.