The Permission Layer

The payment problem for agent commerce is nearly solved.

Stripe Payment Links work. The x402 micropayment protocol has working implementations. AWS has started shipping AgentCore Payments infrastructure. The basic rails exist: an agent can have a payment method, trigger a checkout, and receive a receipt.

What is not solved is the permission problem.

The Permission Problem

A payment proves a payment method was used. It does not prove the human behind that payment method intended this specific purchase, in this context, at this dollar amount, for this outcome.

The gap is clearest in enterprise scenarios. A company gives an agent a company card with a $5,000 monthly limit. The agent books a cloud data feed. Accounting asks: was this pre-approved? Which budget line? What was the procurement justification? Was this in the agent's operating mandate?

The card was charged. Nobody can answer the other questions.

This is not a payments failure. It is a policy and documentation failure. The infrastructure for the purchase existed. The infrastructure for the authorization did not.

What Authorization Infrastructure Looks Like

For an agent-to-service transaction to be auditable after the fact, several things need to exist at transaction time:

On the vendor side: A machine-readable manifest declaring what can be bought, what it does, what pricing rules apply, and what evidence the buyer receives.

On the buyer side: A spend policy document specifying which vendors the agent may transact with, the per-call and per-period limits, and what the agent must capture to satisfy internal approval.

At the point of sale: A structured receipt binding the specific service, the specific principal authorization reference, the dollar amount, and a timestamp the ledger can index.

None of this requires a new payment rail. It requires a documentation standard and the tooling to produce and verify documents against it.

Where Attest Sits

Attest generates the vendor-side half: the machine-readable manifest, the policy framing, and the structured receipt template that makes a service agent-ready.

The readiness report scores a service on whether these elements are in place and what would need to change before an agent could purchase from it with confidence. The manifest and policy packs produce the artifacts directly.

This is narrow by design. The payment rail is not our problem; the authorization trail is.

As agent spending scales from individual experimentation into enterprise operations, the question "can the agent pay?" will matter less than "can the vendor prove the purchase was authorized?" The ledger beats the balance.

---

Attest is the proof layer for agent commerce. Readiness reports, manifests, and policy packs available at getattest.com.au.