We aim to automate coupling proofs for general probabilistic programs with potentially different structures beyond the scope of existing techniques. We define a joint automaton of the two programs of interest, develop checking procedures for the validity of the automaton, and show the soundness of our approach: the existence of a valid automaton implies the desired coupling proof. In the future, we plan to automate the construction of the automaton through a data-driven learning method that constructs the automaton through program traces.