Summary Reflection tomography is often run with beam migration for iterative velocity refinement. Both are ray-based algorithms: beam migration shoots rays down from the surface; while tomography shoots rays up from RMO picking points to the surface. The former ray tracing is straightforward with initial directions defined by a fan of preset angles; however, the latter is difficult, similar to a two-point ray tracing problem. Historically ray tracing in migration and tomography are decoupled. But in this study, we will establish a relationship between them, so that ray paths computed in beam migration provide ray take-off directions for reflection tomography. This coupling concept dramatically boosts the efficiency and accuracy of standard ray-based reflection tomography.