The payments industry is finally catching up to what we've been seeing for months: AI agents are economic actors now. They're buying API credits, ordering food, purchasing domains, paying for SaaS subscriptions. And they need payment infrastructure built for them — not the same rails designed for humans swiping cards at checkout.
Visa and MasterCard are both building this. We've been following it closely because it directly shapes what we're building at AgentCard. Here's what's happening.
Visa Intelligent Commerce
Visa Intelligent Commerce is Visa's push to enable secure, AI-initiated transactions at scale. Right now — including with AgentCard — agents use raw card numbers. It works, but it's not ideal. Visa Intelligent Commerce introduces tokenized payment credentials scoped to specific contexts. This is a big deal.
How it works
- Tokenized credentials — instead of a 16-digit card number, the agent receives a token that represents the payment method. This token can be scoped to specific merchants, transaction types, or spending limits.
- Merchant-scoped authorization — a token issued for buying cloud credits can't be used to order pizza. The payment authorization is contextually bound.
- Network-level enforcement — spending controls are enforced by Visa's network, not by software. This is the same principle that makes virtual cards secure today, but with finer-grained controls.
Where it stands
Over 100 partners. Hundreds of agent-initiated transactions completed in controlled environments. Visa predicts millions of consumers will use AI agents to complete purchases by the 2026 holiday season.
They're building this with major commerce platforms, AI companies, and payment processors. The goal is agent-initiated transactions that are as secure as human ones — but with controls built specifically for autonomous systems. We think they'll get there.
MasterCard Agent Pay
MasterCard's approach is different but complementary. They're focused on enabling agent transactions within regulated banking frameworks. If you're in enterprise — where compliance, audit trails, and institutional controls are non-negotiable — this matters a lot.
How it works
- Bank-integrated agent authentication — agents transact through established banking relationships, not through standalone card numbers
- Regulated transaction processing — every agent-initiated payment flows through the same regulatory framework as human transactions, ensuring compliance
- Scalable agent identity — MasterCard is building infrastructure for agents to have verifiable identities within the payment network, enabling trust at scale
Where it stands
In March 2026, MasterCard and Santander completed Europe's first live end-to-end payment executed entirely by an AI agent within a regulated banking framework. Not a demo. A real transaction, through real banking infrastructure. That's a milestone worth paying attention to.
Fiserv has also integrated MasterCard Agent Pay into its merchant platform, bringing agent-initiated commerce to its network of merchants and financial institutions.
What this means for the market
Here's the reality today: giving an AI agent payment access means sharing card credentials — number, CVV, expiry. The agent fills in checkout forms just like a human would. We built AgentCard on this model because it works right now. But it has clear limitations:
- Card numbers are static — once shared, they can't be scoped to specific uses
- No native agent identity — the payment network doesn't know or care that an agent made the purchase
- Limited merchant controls — you can't restrict a card to specific merchant categories at the network level (without issuer-level controls)
Visa Intelligent Commerce and MasterCard Agent Pay solve all of these. Tokenized, scoped credentials with network-level enforcement, agent identity, and native integration with payment processing infrastructure.
What this means for AgentCard
We built AgentCard to solve one problem: how does my AI agent pay for things? Today, the answer is virtual cards. Your agent creates a card with a fixed balance, grabs the details, and makes purchases. The balance is the spending limit. Simple.
Virtual cards will remain the foundation — they work today, with every agent, at every merchant. But as these new protocols roll out, we're layering them on to offer:
- Tokenized credentials — agents get scoped tokens instead of raw card numbers
- Merchant-level controls — restrict agent spending to specific merchant categories
- Native agent identity — your agent transacts with a verifiable identity, not just a card number
- Network-enforced policies — spending rules enforced by Visa and MasterCard, not just by card balances
We're in active conversations with both Visa and MasterCard on these integrations. When they're ready, AgentCard users will be first to get them.
The timeline
Things are moving fast. Here's roughly how we see it playing out:
- Now — Virtual cards (like AgentCard) are the practical solution for agent payments
- Mid 2026 — Visa Intelligent Commerce and MasterCard Agent Pay reach broader availability
- Late 2026 — Visa expects millions of agent-initiated transactions; MasterCard expanding Agent Pay globally
- 2027 and beyond — Native agentic payment rails become the standard, with virtual cards as a bridge for backward compatibility
McKinsey estimates agentic commerce will generate $3 trillion to $5 trillion in global revenue by 2030. That's not a typo. The payment infrastructure being built right now will power this entire shift.
Getting started today
You don't need to wait for any of this. AgentCard gives your AI agent a virtual card right now — takes about 60 seconds:
npm install -g agentcard
agentcard signup --email you@example.com
agentcard create --amount 25
agentcard details <card-id>
Start with virtual cards today. When native agentic protocols go live, we'll upgrade you automatically.