Evidence for agent-initiated payments
AI agents are about to start moving money on their own. Someone has to be able to prove they did it right.
Pehtee-Nu is building verifiable audit trails for agentic payments: the missing piece between autonomous execution and institutional trust.
The problem
What happens when agents start acting, not just advising
As AI agents take on more of the actual work of moving money (approving vendors, executing payments, reconciling accounts), the question stops being whether they can act and becomes whether anyone can prove they acted correctly.
That's the layer Pehtee-Nu is building: the record of what an agent did, and the reasoning behind it, ready the moment someone needs to ask.
Two different questions
A payment can match its instruction and still be wrong
Did the payment match the instruction?
Card networks have started checking this at the rail itself, comparing what was executed against what was authorized.
Was the instruction sound, and can you show how the agent reached it?
An agent that pays a fraudulent invoice exactly as instructed passes the first check cleanly. This second question is the one Pehtee-Nu is built around.
How it works In development
From the logs you already keep to evidence someone else can check
Pehtee-Nu works after the fact, from the agent and payment records you already keep, and never sits in the payment path. For a specific payment, it rebuilds what the agent did and why.
Who it's for
Built for the people who answer for a payment
Payment operations
Reconstruct what happened on a payment that went wrong, and why.
Payment operations →Treasury
Show what your agents did with company money, one payment at a time.
Treasury →Fraud & risk
Tell a payment that matched its instruction apart from one whose instruction was sound.
Fraud & risk →Compliance & audit
Answer for a specific payment, not only for the policy that was meant to govern it.
Compliance & audit →Research
The timing isn't incidental
Independent research and regulatory signals point the same way at once. We publish what we check, and where we checked it.
The oversight gap, verified
Every claim checked directly against a primary source: the supervisory letters, the framework documents, the sector's own vocabulary.
Read the finding →What's settled, and what's still being tested In development
A dated record of what our research has established and what remains open.
The institutions that get this right first won't be the ones with the fastest agents
They'll be the ones who can prove exactly what happened and why, not just that a policy existed to govern it. If your team moves money with agents, or receives payments that agents send, we'd like to hear how you handle it today.