Models / Mjölnir
Mjölnir Judge 1.0
Checks it, weighs the risk, makes the call
Before your AI agent makes a move, Mjölnir Judge answers three questions at once — can we understand what it is about to do, how risky is it, and are we allowed to let it run — and turns them into one sentence anyone on your team can read. It never makes up its own rules: it can only approve what your rules already approve, and a low risk score never overrides a block. When it cannot be sure, nothing goes forward without a human.
Where it helps
Fail-closedIf it cannot be certain, it holds back — and says exactly what is missing
UtilityCan this action be checked, how risky is it, and may it run without a human?
One rulingEverything it looked at comes back as one plain-English sentence
Required · one call per line — multiple lines are judged as a plan.
Judge an action to see the ruling