What Are Agentic Tokens?
Agentic tokens are card-network payment tokens issued for use by a specific AI agent. A registered agent presents the token in place of the card number to pay on a cardholder's behalf, and merchants, networks and issuers can recognize the payment as agent-initiated. Agentic Tokens is the name Mastercard uses for the credentials in Mastercard Agent Pay; Visa issues tokenized credentials for agents under Visa Intelligent Commerce, and Stripe calls both "agentic network tokens." A token carries the agent's identity and per-transaction controls, while totals across many purchases are tracked separately.
Updated
What are agentic tokens?
They extend network tokenization. In EMVCo's description, payment tokenisation replaces the card number with "a unique alternative value, the EMV Payment Token," and "an EMV Payment Token is constrained in how it can be used. For example, to a specific merchant, device or payment scenario." An agentic token adds the agent to those constraints.
Mastercard introduced Agentic Tokens with Agent Pay on April 29, 2025, building on the tokenization it already uses for contactless mobile payments and cards on file. In September 2025 it said that "each time an agent acts on your behalf — even for a recurring payment — it does so with the permissions and limits you define." Visa's Intelligent Commerce launch describes AI-Ready Cards that replace "card details with tokenized digital credentials" and "confirm that a consumer's chosen agent is allowed to act on a consumer's behalf." Stripe, which provisions both behind its Shared Payment Tokens, says agentic network tokens "allow authorized AI agents to initiate payments on a customer's behalf without exposing underlying card details." Both programs are covered in Mastercard Agent Pay and Visa Intelligent Commerce, and the checkout protocols they plug into in UCP vs ACP.
How agentic tokens work
Mastercard's Agent Pay Acceptance Framework (October 14, 2025) describes steps 1, 3 and 4 below; steps 2 and 5 are general network-token mechanics:
- 1.The agent is registered. The framework "begins by registering and verifying AI agents before they are permitted to transact on the Mastercard network. Each agent is uniquely identified and then enabled to initiate transactions using agentic tokens."
- 2.The cardholder enables a card for the agent. As with any network token, a token is provisioned for that use, here bound to the agent with the cardholder's permissions and limits.
- 3.The merchant verifies the agent, for example with Web Bot Auth "at the Content Delivery Network (CDN) layer."
- 4.The agent pays. It can submit "a Dynamic Token Verification Code — an agentic token formatted for standard card payment fields" through an existing checkout form.
- 5.The issuer authorizes. As with any network token, the network maps the token back to the card number before the issuer decides. Visa said in June 2026 it is enriching token data with "the transaction type, where the token is being used and who is making the payment," plus a "token assurance signal."
What the token answers, and what it leaves open:
| Question at authorization | Answered by |
|---|---|
| Is this credential valid, and bound to this agent? | The network, from the token |
| Is this purchase within the token's transaction-level controls? | The network and issuer |
| Has the mandate's total budget been used up? | The network, when a Verifiable Intent budget constraint is attached |
| Does it fit the platform's budget across agents, users and cards? | The agent platform |
| Has this purchase already been submitted? | The agent platform |
Mastercard says agentic tokens "enable programmable transaction-level controls." Those compare one purchase with its rules and need no history; a budget across purchases does. The draft Verifiable Intent constraints spec, maintained by Mastercard, states that "Payment networks MUST maintain stateful tracking of cumulative spend per L2 mandate pair when a mandate.payment.budget constraint is present." The bottom two rows are the platform's authorization-time context: remaining budget net of purchases in flight, per-agent limits and what was already submitted, read before the agent presents the token.
Why agentic tokens matter in real-time systems
A token is scoped to one agent and one card. Spending rarely is. An illustrative household runs two assistants, each holding a token on the same card, under one weekly grocery budget set in the assistant app. Each purchase passes its token's controls; together they exceed the budget, and no single token saw both. Procurement agents spending across many employees' cards hit the same wall.
Only the platform sees all of it. Tacnode does not touch the token or the authorization, and enforces nothing inside a network's or issuer's ledger. It can hold the platform's side: budgets, reservations and purchase intents held in Tacnode are transactional, so two agents cannot both reserve the last of a shared budget, and authorization results are written back as context for the next purchase. The buyer's side of agent checkout is covered in how shopping agents buy.
FAQ
Related terms
An agent wallet lets an AI agent hold or spend money on its own within set limits. Its checks cover one wallet; budgets shared across agents live elsewhere.
An agent mandate is a signed record of what a user authorized an AI agent to buy: one exact purchase, or constraints such as a budget the agent must keep to.
Authorization-time context is everything a payment or purchase decision must read when it authorizes: mandate status, remaining budget, limits and recent spend.
Agent spend limits cap what an AI agent may commit per transaction, period or agent. They are only as accurate as the spend total they are checked against.
Mandate repudiation names a dispute in which a user denies a purchase an AI agent made within the authority they gave it. A decision-time record settles it.
Velocity checks are fraud rules that count or sum activity per card, device, IP address or account over a time window, and act when a threshold is crossed.