← Go back

Guide // MandateGuard

How do merchants verify an agent's receipt against an intent?

By the Ingenious AI team · Published 2026-11-14 · Reviewed 2026-11-14

▸ Run the free live demo See pricing

Free covers 1 agent and 1 policy with the full authorize flow, signed receipts and the approval inbox — no card required to start.

Short answer

A merchant verifies an agent's receipt against an intent by calling MandateGuard's verify-receipt check with its read-only merchant key, the intent_id and the receipt ID. The answer reports four component checks — the receipt's hash recomputes, its signature is valid, it sits correctly in the workspace hash chain, and it is linked to that intent — plus an overall verdict. Run it before fulfilment, store the result on the order, and treat any failed component as a stop-and-classify event: linkage failures are usually clerical, hash or chain failures mean the order must not be honoured on that receipt.

What does receipt verification check?

Receipt verification is the merchant-side check that a claimed mandate-backed purchase really happened under a registered intent. Using a merchant role key, the merchant calls the verify-receipt path for an intent and receipt ID; MandateGuard returns whether the receipt is verified and the result of each component check: receipt hash, signature, membership in the workspace's hash chain, and linkage to the named intent. The merchant key can perform this check but cannot alter the intent or the receipt.

The merchant's dilemma

An agent presents an order and claims a mandate covers it. The merchant wants the business and fears the chargeback: if the mandate was unreal, exhausted or already revoked, the merchant may ship goods against authority that never existed. Asking the agent is useless — the agent is the claimant. Asking the issuer by phone is slow and unscalable. The verify-receipt check resolves the dilemma mechanically: the merchant asks the registry itself, using its own read-only key, and receives a verdict grounded in the hash-chained record rather than in anyone's say-so. Verification converts trust-me into check-for-yourself, which is the only basis on which mandate-backed commerce scales.

The four checks, in plain terms

Hash: the receipt's stored content recomputes to its recorded hash, so the receipt was not edited. Signature: the receipt carries the server's valid HMAC signature, so it was genuinely issued by MandateGuard rather than fabricated. Chain: the receipt sits correctly in its workspace's hash chain, linked to predecessor and successor, so it was not inserted into history after the fact. Linkage: the receipt is recorded against the named intent, so this purchase was made under this mandate — not under a different intent or as ordinary spend. A full pass means the claim is exactly what it appears to be. Each failure mode means something specific and different, which is why the check reports components instead of a bare yes or no.

When to run the check

Run it at the moment of commitment: before fulfilment, before provisioning, before extending mandate-backed terms. Verification after shipment is forensics. The check is a single call (and is also available as an MCP tool and in the registry UI for manual desks), so cost is no argument for skipping it on larger orders; many merchants set a value threshold below which streamlined processing applies and above which verification is mandatory — mirroring, on the receiving side, the approval thresholds issuers use on the spending side. Store the verification result with the order record: if a dispute arrives months later, the dated verdict is your evidence that diligence happened at the right time.

Reading failures without panic

A failed check is information, not accusation. A linkage failure often means the buyer cited the wrong intent — a correctable clerical issue resolvable with the issuer. A hash or chain failure is more serious: the record itself does not recompute, and the order should not be honoured on that receipt while the issuer investigates. An unknown receipt ID may mean the purchase ran under a plain policy rather than an intent, in which case it may still be a legitimate order — just not a mandate-backed one, and it should be processed (or not) on ordinary terms. The discipline is to stop, classify, and contact the issuer with the specific failed component, which focuses their response immediately.

Verification and intent state

Verification answers whether a receipt is genuine and linked; the intent's current state answers a neighbouring question — is this mandate still live, and how much budget remains? Merchants should read both at order time. A genuine receipt against a since-revoked intent usually represents a purchase made while the mandate was alive, which the intent's event chain can confirm by sequence: the authorisation precedes the revocation. That ordering is precisely what the chained event log is for. Honour and payment decisions remain commercial judgements, but they become judgements made over a verified timeline instead of competing recollections.

Building verification into the relationship

Verification works best when it is set up before it is needed: keys exchanged at onboarding, the intent_id written into the order flow, the check rehearsed with a known-good receipt, and the failure contacts agreed (who at the issuer receives a failed-check report, through what channel, with what response expectation). Merchants who integrate verification report a side benefit beyond fraud prevention: mandate-backed orders become their cleanest paperwork, because every order carries its authority with it. In a market where agent purchasing is new, that cleanliness is a competitive feature — the merchant who can verify is the merchant issuers route mandates to.

Five actions to take now

  1. Give the merchant a labelled merchant-role key at onboarding
  2. Verify at order time, not at dispute time — the check is cheap
  3. Check all four results: hash, signature, chain and intent linkage
  4. Treat a failed check as a stop: contact the issuer before honouring the order
  5. Keep already-verified receipts honoured if an intent is later suspended

Worked example: verified once, caught once

A merchant receives a mandate-backed order for 310 dollars referencing intent and receipt IDs. Its order system calls verify-receipt before fulfilment: verified, with hash, signature, chain and linkage all passing, and the intent showing 690 dollars remaining of its 1,000-dollar monthly-scoped budget after this purchase. The order ships the same day, with the verification result stored on the order record.

A week later a second order arrives claiming the same receipt ID against a different, larger amount. Verification fails: the receipt's recorded facts do not match the claim. Fulfilment stops, the merchant contacts the issuer with both order references, and the episode resolves as a reseller's clerical fraud rather than a shipped loss. The check cost seconds; the dispute it prevented would have cost weeks.

Frequently asked questions

What do I need to verify a receipt?

A merchant (or agent) role key, the intent_id and the receipt ID. The issuer provides the key and the intent reference at onboarding.

What if the intent was suspended after the purchase?

Verification checks the receipt against the intent's record, including its chain of events. Honour decisions for already-verified purchases are a commercial matter; the verification result tells you the purchase genuinely occurred under the intent.

Can verification be done in the registry UI?

The registry includes a third-party verification panel where intent and receipt IDs can be checked interactively — useful for manual order desks.

What does a linkage failure mean?

The receipt exists but is not linked to that intent — it may belong to a different mandate or to ordinary policy spend. Do not honour it as intent-backed.

Is my merchant key able to spend or change anything?

No. Merchant keys are read and verify only, by design.

Keep reading

Put the policy in force: spend envelopes · intent registry · approval inbox · how the proof works · all MandateGuard guides