MOLTAP is an automated prover of modal logic that is not just of theoretical interest. There are three aspects that make it useful as a practical tool.
First of all the logic behind the prover has been shown to be sound and complete. This means that the results will be as expected. A tool that could spout nonsense is useless.
The second quality that makes MOLTAP useful is the fact that it has got a decent user interface. This is the point where other systems fail. MOLTAP includes both a web interface for easily entering formulas and getting direct results, as well as a command line program for conveniently working with larger formulas.
Finally the ability to extract counter models can be very useful. It can be used as an educational tool to teach students about Kripke models and validity of formulas. Counter models can also be used for quickly finding a model with a given property. For example to find a model in which x ∧ M1¬K2 x holds simply enter the negation into the prover, ¬(x ∧ M1 ¬K2 x).
There are still several tasks remaining. The most important of these is the correct formulation of rules for group and common knowledge.
The rules for group knowledge in system K are still understandable, but when different systems are used the situation becomes exceedingly difficult.
Similarly assuming common knowledge is not too hard, a formula like □* φ → ψ can simply be interpreted as (φ ∧ □ φ ∧ □□ φ ∧ ...) → ψ. But this interpretation already leads to problems with non-termination in rare corner cases such as ¬□*◇φ. The program can keep adding boxes because the diamond rule creates new worlds to match. It turns out that in system S5 this does not happen, and MOLTAP does implement such a rule for common knowledge.
Proving common knowledge on the other hand is very difficult. Usually induction rules are required, which again have a large risk of non-termination.
Another avenue for further work is mixed systems. In such a system one set of relations can be for example full equivalence relations, while another set is only serial. It is likely that this opens up many interesting options for modeling real world problems.
Finally it would be nice if a derivation could be shown for valid formulas. This would convince the user of the correctness, and allow the output to be used beyond the program. It is not hard to keep track of which rules are applied and upon reaching an axiom this list can be reversed to give a derivation. It might also be useful to show derivations in a more traditional proof system, which the user is more likely to be familiar with.