Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)

OpenAI Launches Wallet: Dot Pays for Purchases with a Token, Not a Card Number

OpenAI has launched Wallet, a payment wallet through which the Dot agent can make purchases on its own. The user adds a card once, and the agent receives a payment token instead of the actual card details. Every purchase must be confirmed with Face ID or a fingerprint, so the final say stays with the human. For now, the feature is available only to Pro subscribers in the US, which means developers building agents should start designing a "token plus confirmation" scheme now.

AI-processed from TG: @TochkiNadAI (Точки над ИИ); edited by Hamidun News
OpenAI Launches Wallet: Dot Pays for Purchases with a Token, Not a Card Number
Source: TG: @TochkiNadAI (Точки над ИИ). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @TochkiNadAI (Точки над ИИ) ↗
Editor-in-Chief
Key metric #1The agent sees only a payment token; the card number is never passed to it
Key metric #2Access is currently limited to Pro subscribers in the US
Key metric #3Limits and purchase categories without confirmation are promised next

Agents have long been able to read email, search for products, and correspond with support teams. But at the last step they ran into the same problem: to buy something, someone has to bring in a human with a wallet. OpenAI has addressed this for its Dot agent and released Wallet, a payment wallet through which the agent can pay for purchases on its own, but with the owner's oversight.

How Wallet works

The scheme is simple. The user adds a bank card once. Dot receives neither the card number, nor the expiration date, nor the security code; instead, the agent works with a payment token. This is the same principle behind Apple Pay and Google Pay: the merchant or intermediary sees a stand-in record tied to the device or a single transaction, while the real card details stay with the payment network.

For developers, the reason matters more than the technology itself. If an agent is compromised through a malicious web page, or accidentally writes data to a log, it is the token that leaks, not the card. A token can be revoked without reissuing the card itself. This lowers the cost of a mistake, and agents that read other people's pages and emails still make mistakes often.

The second layer of protection is the human. The owner confirms each purchase with Face ID or a fingerprint. The agent can find a product, go through to checkout, and prepare the order, but the money leaves only after biometric approval. It resembles a "confirm payment" notification in a banking app, except that the request comes from your agent.

What's next and who gets access

For now, Wallet is available only to Pro subscribers in the US. The announcement gives no timeline for other countries. OpenAI also promises the next stage: limits and purchase categories within which the agent can pay without confirmation. For example, subscriptions up to a certain amount, or food delivery spending. How exactly these rules will be configured has not been specified.

This next stage will be the real test. Confirming every purchase with Face ID is safe, but it is not autonomy: a person still holds the phone. Autonomy begins where the agent spends money under a predefined policy and the owner learns about the purchase after the fact. This raises questions that have no answers yet: what to do about disputed charges, who is responsible if the agent misunderstood the task, and how to verify that the product matches the request.

What this means in practice

The author of the original post, who has used Dot since launch, takes the novelty calmly. He does not build systems on the agent and does not consider it central: Dot sends him news digests and reminders, monitors his email, and alerts him to important messages. Most often he has Dot communicate with the outside world, mainly with support services. Correspondence sometimes drags on for several days, and the agent handles it well. About the wallet he says briefly: fine, he'll test it when it's offered. Requests to pay for something take him little time, so there is no pressing need to optimize such a simple action.

This is an honest assessment, and it shows the real state of affairs. While confirmation is required for every purchase, the gain from Wallet is modest: you save on entering card details, but not on attention. The real benefit will come with limits and categories, when the agent can handle routine tasks, such as paying small subscriptions or recurring orders.

For those building their own agents, the conclusion is broader than one OpenAI product:

  • Do not give the agent real payment details; use tokens that can be revoked.
  • Keep human confirmation on payments, especially in the early stages.
  • Set limits on amount and category before giving the agent more freedom.
  • Log every payment attempt so that disputed cases can be investigated later.

Wallet is not a breakthrough but a careful engineering step: safety and control first, autonomy later. How quickly OpenAI reaches the second part and opens access outside the US will become clear in the coming months.

Story

Key Questions Answered

How Wallet works?

The scheme is simple. The user adds a bank card once. Dot receives neither the card number, nor the expiration date, nor the security code; instead, the agent works with a payment token. This is the same principle behind Apple Pay and Google Pay: the merchant or intermediary sees a stand-in record tied to the device or a single transaction, while the real card details stay with the payment network. For developers, the reason matters more than the technology itself.

What's next and who gets access?

For now, Wallet is available only to Pro subscribers in the US. The announcement gives no timeline for other countries. OpenAI also promises the next stage: limits and purchase categories within which the agent can pay without confirmation. For example, subscriptions up to a certain amount, or food delivery spending. How exactly these rules will be configured has not been specified. This next stage will be the real test.

What this means in practice?

The author of the original post, who has used Dot since launch, takes the novelty calmly. He does not build systems on the agent and does not consider it central: Dot sends him news digests and reminders, monitors his email, and alerts him to important messages. Most often he has Dot communicate with the outside world, mainly with support services. Correspondence sometimes drags on for several days, and the agent handles it well. About the wallet he says briefly: fine, he'll test it when it's offered.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…