Custodian is the trust layer underneath. The agent can only request — a deterministic kernel decides. It cannot exceed its limit, approve its own escalation, or lie its way past the rules, because the boundary is enforced below the agent at the OS kernel — not promised in a prompt.
Spend caps and approval flows are commodities now. The hard problem isn't limiting a number — it's that the agent can be wrong, or can lie, and that it shouldn't be trusted to route money through an approved path in the first place.
The control lives in their custodial cloud. The agent reaches money by calling their SDK, and safety rests on the assumption it'll use the approved path. They cap the dollar amount — but never check whether what the agent claims is even true.
The control lives in Landlock + kernel egress policy. The agent literally cannot open a socket to a payment endpoint the OS hasn't allowed. And a deterministic verifier checks every fact the agent asserts against ground truth — so it can't lie its way to a payout. Non-custodial, rail-agnostic, self-hosted.
The agent reads the messy real world and makes a recommendation. Then three deterministic, zero-AI layers get the final say — and any one of them can stop the money.
Nemotron reads the email, invoice, or task and proposes an action — refund, payment, provision a GPU. It recommends. It never decides money.
can be wrong · can lieEvery factual claim the agent made is resolved against ground truth. A claim the data refutes is flagged CONTRADICTED before anything downstream trusts it.
deterministic · zero-AIBands and caps decide AUTONOMOUS / ESCALATE / DENY. Over the cap requires a real human signature (Twilio Verify SMS). The agent never holds both keys.
enforced at OS levelThe agent can lie — and money still can't move wrong. When a customer invents a story to get a refund and the AI recommends approve, the verifier catches that the claim is contradicted by the ledger and the kernel overrides the AI. No competitor can demonstrate this, because their model is "agent asks → check the limit," not "agent asks → check if the agent is lying."
A real Nous Hermes agent, in a real kernel sandbox, paying real Stripe PaymentIntents — protecting ArgoBox, a production homelab. These numbers are pulled live from the running system as you read this.
The whole category ships caps, approval, and audit. Only one row below is shared — the bottom four, together, are Custodian's alone.
| Capability | Payman · Skyfire · Catena · Rain · Ramp | Custodian |
|---|---|---|
| Spend caps · approval · audit | ✓ table stakes | ✓ |
| Catches the agent lying (facts vs ground truth) | ✕ | ✓ |
| Enforcement below the agent (kernel, not API) | ✕ | ✓ |
| Non-custodial · rail-agnostic · self-hosted | ✕ they hold funds | ✓ |
| Domain-general — one kernel, many modules | ✕ money only | ✓ |
Watch a real agent recommend approving a fraudulent refund — and watch the deterministic kernel override it, with real Stripe IDs and an append-only audit trail.
Money is just the first module. The same kernel governs any consequential action an AI agent can take — provisioning, payroll, data egress, infrastructure.