What is Visa TAP? The Trusted Agent Protocol, explained

What is Visa TAP?

Visa TAP (Trusted Agent Protocol) is an open framework launched by Visa in October 2025, with input from Adyen, Coinbase, Microsoft, Shopify, Stripe, and others. It gives an approved AI agent a way to prove to a merchant's website that it's a legitimate shopping assistant, not a random bot.

The simplest way to think about it: UCP and ACP are about what the agent says to the store. Visa TAP is about getting the agent through the front door in the first place.

Here's the problem it solves. Merchants have spent years building defenses against bots. Scrapers, scalpers, credential stuffers, fake account farms. Most online stores sit behind bot protection that's designed to spot anything automated and block it. That worked fine when every good customer was a human. Now some of the best customers are agents, and to a bot detector, a helpful shopping assistant and a ticket-scalping script look pretty much the same.

So merchants are stuck. Block all automated traffic and lose real sales. Let it all through and invite fraud. Visa TAP gives them a third option: let the good agents identify themselves. Knowing who the agent is is the first step. Deciding what it should be allowed to do comes right after.

What Visa TAP actually covers

With Visa TAP, every request an approved agent sends to a merchant carries a digital signature, like a stamped badge that can't be faked. That badge lets the agent pass along three things:

1. Who the agent is and what it's here to do: the merchant can tell a registered agent apart from unknown traffic, and see whether it's just browsing or actually trying to pay. A store might happily let agents browse freely but want more checks at checkout.

2. Who the agent is shopping for: the agent can link itself to the customer it represents, so a returning customer can still be recognized, even when they're not the one clicking.

3. How the customer intends to pay: the agent can share payment details in a way the merchant can trust came from the same verified agent.

The badge also can't be reused. Each one is tied to a specific website, a specific moment, and a specific page, so a criminal can't copy a good agent's badge and replay it somewhere else. What the badge doesn't carry is a verdict. It tells the merchant who the agent is, not whether this particular request makes sense.

Why it works at the front door, not the checkout

Most of the other protocols kick in once the agent is already shopping. UCP organizes the cart. ACP and AP2 handle the payment and the permission. Visa TAP works earlier, at the very first moment the agent knocks on the website.

That placement is the whole idea. If a good agent gets blocked at the door, it never reaches the cart, and nothing else matters. If a bad bot gets through the door, it can scrape prices, hoard inventory, or test stolen cards long before it ever tries to pay.

Think of it like a building with a lobby. UCP, ACP, and AP2 are what happens upstairs: picking the product, signing the receipt. Visa TAP is the badge reader in the lobby. It doesn't decide what you buy. It decides whether you get in the elevator, and it tells the people upstairs who just came in. What that visitor is allowed to do once upstairs is a separate decision.

It also plays well with the others. An agent can carry a Visa TAP badge to get through the door and then shop using UCP or ACP once it's inside.

How a merchant adopts Visa TAP

For most merchants, the answer is: your security provider does the heavy lifting.

Visa TAP was designed to work with the web infrastructure merchants already have. There's no need to rebuild the checkout page. The bot protection and security providers that already sit in front of most online stores can check the agent's badge on the merchant's behalf and offer ready-made rules that let verified agents through. The merchant decides the policy (for example, "let verified agents browse, but only let them check out if they're linked to a known customer"), and the provider enforces it. Rules like that are only as good as the information behind them, which is why merchants increasingly pair Visa TAP with a layer that looks at what each verified agent is actually trying to do.

On the agent side, developers building agents can sign their requests the Visa TAP way, so their agents can shop across those merchants without being mistaken for bots.

Merchants with a different security setup can verify the badges themselves. Visa publishes the spec and sample code openly through its developer center and on GitHub.

Where Beltic fits in

Visa TAP tells the merchant who walked in. It doesn't tell the merchant what that visitor is about to do.

Go back to the badge reader in the lobby. A valid badge means the person works in the building. It doesn't mean they should be walking out with three laptops. A verified agent is the same. It's a real, registered shopping assistant, and it might still be acting on a request that makes no sense for this customer, or on behalf of an account that was taken over last night.

That's the gap Beltic fills. Visa TAP answers "is this a known agent?" Beltic answers the next question: "should I let this known agent do this specific thing?"

Here's how it works in practice. A verified agent gets through the door and starts shopping. Beltic takes everything the merchant already knows, including the Visa TAP badge, the customer it's linked to, and what the agent is trying to buy, and looks at whether the request is clear, whether it matches what this customer normally does, and how risky it is. That becomes a clear risk read, and the merchant's own rules decide: let the purchase go through, stop it, or ask the customer to confirm.

Visa TAP and Beltic work better together. A verified badge is one of the strongest signals Beltic can use, and Beltic is what turns that signal into a decision. And because Beltic works the same way across every protocol, a merchant that relies on Visa TAP today doesn't have to start over when agents arrive through UCP, ACP, or AP2 tomorrow.

For the merchant, the payoff is being able to open the door to agents without opening the door to chargebacks. Good agents get a smooth path. Verified agents doing unusual things get a second look.

Visa TAP checks the badge at the door. Beltic watches what happens once the agent is inside.

This post is part of our series on agentic protocols. For how they all fit together, read What are agentic protocols? A plain-English guide.