What Is an Intent Object? How AI Agents Get Authorization to Spend
Key takeaways
- An intent object is a structured, line-item purchase contract that a user approves before an AI agent executes anything.
- It works as a receipt before the purchase: the user sees the merchant, items, quantities, total, and fulfillment details in advance.
- Its value comes from enforcement. The system should check the approved terms throughout execution, not once at the start.
- An intent object is more specific than a card spend limit: it scopes the merchant, the items, and the price, not only an amount and a category.
- Similar records exist elsewhere in the industry, such as the checkout and payment mandates in the Agent Payments Protocol (AP2).
An intent object is a structured, line-item purchase contract that a user approves before an AI agent executes it. It names the merchant, the specific items, the quantities, the price ceiling, and the fulfillment details. It gives the agent permission to complete one bounded purchase, not open-ended access to a card.
Why does an agent purchase need an intent object at all?
An agent purchase needs an intent object because a chat message is not an authorization record. If the user types "yeah, order it" and the agent charges a card, the record has no amount, no merchant, no items, and no way to show later what the user believed they agreed to.
That gap causes three problems. The user cannot see what will happen before it happens. The builder cannot prove what was approved when a dispute arrives. And nothing constrains the agent if the merchant's page costs more, shows a different item, or adds fees. An intent object addresses all three, because the approval becomes a structured record instead of a conversational turn.
What is inside an intent object?
An intent object contains the answer to each question a user would want answered before saying yes: where, what, how many, how much, and when.
| Field | What it pins down |
|---|---|
| Merchant | The specific store where the purchase will happen, not a category |
| Line items | Each product, by name and variant, so a substitution is visible |
| Quantities | How many of each, so an agent cannot increase volume |
| Price ceiling | The maximum total the user agrees to, including fees |
| Fulfillment | Delivery or pickup details, address, and timing where relevant |
| Scope and expiry | How long the authorization is valid and what it covers |
An intent object is a receipt that the user sees before the purchase happens.
A receipt is the document users already trust to describe a transaction precisely. An intent object carries the same detail, shown before money moves, at the moment the user can still say no.
How does approval work in practice?
In practice, approval of an intent object follows five steps, from the user's request to a confirmed purchase or a stop.
- 1
The user states what they want: The user asks the agent to order dinner, buy a specific item, or book a table.
- 2
The system drafts the intent object: The request becomes a structured contract with the merchant, items, quantities, and ceiling set to concrete values.
- 3
The user approves it: A scoped prompt shows exactly what will happen, and the user confirms with a biometric or passkey, as with an Apple Pay payment sheet. The prompt appears inside the product the user already uses, with no redirect.
- 4
The agent executes within the terms: The purchase runs at the merchant, and the system checks each step against the approved terms.
- 5
The purchase completes or stops: If the merchant's terms differ from the approved terms, execution stops instead of continuing on the user's card.
What stops an agent from spending outside the approved terms?
Enforcement by the infrastructure stops an agent from spending outside the approved terms, not trust in the agent. A well-designed system checks the approved terms throughout execution, instead of once at the start.
The cases this covers are the ones that occur in real checkouts. A merchant raises the price between selection and checkout. The store substitutes an unavailable item. A fee, a surcharge, or a default tip pushes the total above what the user saw. In each case, the system stops the transaction or asks for a new approval, instead of completing it at a total the user did not agree to.
A prompt is an instruction that an agent can reinterpret. An object is a constraint that the infrastructure applies. That difference is the reason to model intent as an object.
Enforcement can also reach the payment credential. A credential issued for one approved purchase can carry the approved ceiling, for example as a limit on a network token or a single-use card. Then the limit is a property of the credential, not a check the agent must run.
How is an intent object different from a card spend limit?
An intent object scopes the specific merchant, items, and price of one purchase, while a card spend limit sees only an amount, a merchant category, and sometimes a merchant identifier. A spend limit cannot see what is in the cart.
- Amount, for a spend limit: It can block a charge over $80. It cannot tell whether that $80 bought the item the user asked for.
- Category, for a spend limit: It can allow restaurants. It cannot restrict the purchase to the one restaurant the user named.
- Merchant, for an intent object: Approval is scoped to the actual store, so an agent cannot move the same budget to a different store.
- Items, for an intent object: Approval is scoped to specific products and quantities, which is the level at which users think about a purchase.
- Price, for an intent object: The price is a ceiling on a known basket, not a standalone number, so a substitution or a surcharge is detectable.
The two controls work together. A card control is a broad outer boundary. An intent object is the specific contract for one purchase inside it, and it matches what the user thinks they authorized.
How does an intent object relate to an AP2 mandate?
An intent object and an AP2 mandate are both records of what a user authorized an agent to do. The Agent Payments Protocol (AP2), which Google created, defines signed records called mandates. An open checkout mandate records the user's constraints and goals before the cart is final, and a closed checkout mandate records authorization for a specific, finalized checkout. The concept is the same: the approval becomes a structured, checkable record that travels with the purchase.
Why do issuers and merchants benefit from intent-level data?
Issuers and merchants benefit from intent-level data because it gives them more context to approve a legitimate agent purchase. Agent-initiated purchases can look anomalous: the device is unfamiliar, the pattern is unusual, and the issuer often has little data about who started the purchase. With thin context and unusual signals, a decline is the low-risk decision.
Intent-level data changes that input. When authorization data from an approved intent object reaches the transaction, the decision can include the approved merchant, the approved basket, the approved ceiling, and the fact that a specific person confirmed it with a biometric or passkey. Network programs such as Visa Intelligent Commerce and Mastercard Agent Pay move in this direction, so that agent-initiated payments carry more context.
The same record supports the user. When an agent buys with the user's own card under a clear authorization, the purchase is a cardholder-authorized transaction, and the user keeps the normal dispute rights of that card.
Where does the intent object fit in the agent payment stack?
The intent object is the consent layer of the agent payment stack, and the other layers depend on it. The stack has four layers: verified agent identity, intent and consent, credentials, and execution. Identity shows that the agent is accountable. Credentials make the purchase chargeable without exposing a card number. Execution completes the checkout. The intent object sets the limits for all of them.
The credential was never the hard part of agent payments. The hard part is knowing precisely what a person authorized and enforcing it while software acts for them.
Frequently asked questions
What is an intent object in agentic commerce?
An intent object is a structured, line-item purchase contract that a user approves before an AI agent executes it. It specifies the merchant, items, quantities, price ceiling, and fulfillment details, and it gives the agent permission for one bounded purchase instead of open-ended access to a payment credential.
How does an intent object stop an agent from overspending?
An intent object stops overspending when the system enforces its terms throughout execution. If the merchant raises the price, substitutes an item, or adds an unexpected fee, execution stops instead of completing at a total the user did not approve.
Is an intent object the same as a card spending limit?
No. A card spending limit sees an amount and a merchant category. An intent object scopes the purchase by specific merchant, line items, quantities, and a ceiling on a known basket. The two work together: the card control is a broad outer boundary, and the intent object is the contract for one purchase inside it.
Why is an intent object called a receipt before the purchase happens?
An intent object is called a receipt before the purchase because it contains the same precise detail as a receipt, shown before money moves. The user sees the merchant, the line items, and the total while they can still decline, so they approve a specific transaction instead of general permission to spend.
Is an intent object the same as an AP2 mandate?
An intent object and an AP2 mandate are similar concepts. Both record what a user authorized an agent to do. AP2, the Agent Payments Protocol that Google created, defines checkout mandates: an open mandate records the user's constraints, and a closed mandate records authorization for a specific, finalized checkout.
Do intent objects help transactions get approved?
Intent objects can improve authorization context. Agent purchases often look anomalous to issuers because the context is thin. When data from an approved intent object reaches the transaction, with the user's biometric or passkey confirmation, issuers and merchants get a better signal to tell legitimate purchases from fraud.
Related articles