I find it a bit counterintuitive that I can't query such a system interactively and ask questions of it about a proof. Is it safe to say that the architecture

I find it a bit counterintuitive that I can't query such a system interactively and ask questions of it about a proof. Is it safe to say that the architecture in this paper captures MML syntax and the "bounds" within which the software operates?