Audit is the product

Give every agent a chain of command. That line only means something if the chain leaves a mark: a human authorizes, policy enforces, and the record of what happened is the thing you can actually inspect afterward.

Security tooling often treats the audit trail as exhaust — the thing kept around in case something goes wrong. We start from the opposite premise. When an agent acts on someone’s behalf, the record of that action is not a side effect; it is part of the result. Govna commits an audit intent before a connector call is dispatched, and if that commit cannot be made, the call is denied rather than allowed to run un-audited (AUD-001).

A call that cannot be receipted is a call that does not happen.

That fail-closed rule only holds together if the receipt itself can be trusted end to end. Every dispatched call carries a correlation identifier threaded from intent through completion, so a receipt can be traced back to the exact session, identity, and tool that produced it, not reconstructed after the fact from logs that may not agree with each other (AUD-002).

Evidence makes the system examinable. Govna shows the mechanism directly: humans authorize, policy enforces, evidence proves. Read the security brief, click through to the underlying proof record, and inspect the chain of command for yourself.