Levee / Agentic payments glossary
Agentic payments glossary
The terms you'll meet when AI agents start spending money, defined in plain English.
- Agentic payments
- Payments initiated and completed by AI agents on behalf of a person or company, using cards, bank rails or stablecoins.
- A control that decides at the moment of payment whether an AI agent's purchase matches what a human approved, returning approve, decline or challenge.
- Spend mandate
- A human-approved, signed instruction that defines what an AI agent may buy: merchants, purpose, amount and time window. In Levee, mandates are passkey-signed.
- Declaration
- The agent's own statement of what it is buying and why, evaluated alongside the mandate and policy at authorization time.
- Enforcement point
- The place in the payment flow where a decision is enforced. For cards, this is the issuer's real-time authorization path, outside the agent.
- Fail closed
- A safety property: if a decision can't be made in time, the payment is declined rather than allowed.
- Challenge
- A decision outcome that sends the purchase back to a human approver for a quick yes or no instead of approving or declining automatically.
- Strictest wins
- A rule for combining checks: when the policy, mandate and other signals disagree, the most restrictive outcome is applied.
- Evidence ledger
- A tamper-evident, hash-chained record of every authorization decision and its reason, used for chargebacks, finance review and audits.
- Evidence pack
- An exportable bundle of the records behind a decision, prepared for a dispute, audit or finance review.
- A real-time callback from a card issuing platform that lets an external service approve or decline a card transaction before it completes.
- MCC (merchant category code)
- A four-digit code that classifies a merchant's business type. MCC blocks restrict spending by category but can't express purpose.
- x402
- An open approach to internet-native payments built on the HTTP 402 Payment Required status code, letting software and AI agents pay for resources programmatically.
- Prompt injection
- An attack where malicious text manipulates an AI agent into taking actions its user didn't intend, including making unauthorized purchases.
Related
Levee is running a free design partner pilot for platforms putting AI agents in charge of spend.
Become a design partner