Paying a supplier used to mean bank details, a form and a wait. Stripe now lets you find a company by its Stripe Profile handle and send it money from the Dashboard or the command line. An agent can run that command line too.
Typing a handle takes a second. Write down who it's allowed to be before your agent types one.
What Stripe launched
On Oct 27, Stripe announced Stripe Pay on X and LinkedIn. Per PYMNTS, the post read: "Find companies using their Stripe profile handle and send funds for free... Available in the Dashboard and CLI." RuntimeWire and The Paypers covered it the same way. "Free" and "instant" are Stripe's words.
The CLI reference fills in the mechanics. The sender needs a funded financial account. The recipient needs a financial account linked to its Stripe profile, and you name it by @handle or by its profile_ network ID. A --dry-run flag checks the transfer without sending it.
Stripe also published an agent skill for Stripe Pay. Its safety rules tell an agent to send nothing unless asked, show the exact command, review first, skip the --yes flag until the user confirms, never guess a recipient, and never print keys. It also says the command "reports the fee and delivery timing before confirmation. Don't assume that a transfer is free or instant." Read that line next to the launch post. The cost and timing you see in review are the ones that count.
Patrick Collison put the stakes plainly: "A2A payments will be more important than P2P payments," per PYMNTS. That's his view.
The pay rule, in five steps
Steps 1, 4 and 5 are our suggestions. Stripe's docs don't cover them. Steps 2 and 3 follow Stripe's own rules.
1. Write a one page pay rule
Four lines is enough:
- Who it can pay. An allowlist of handles, each one tied to a supplier you already know.
- How much. A cap per send and a cap per week.
- From where. The one financial account it draws on.
- Who confirms. The named person who signs off on each send.
2. Check the handle, never search and send
Stripe's docs say not to guess the recipient. Use Stripe Directory, then have a person pick the exact business. Before the first send to any new handle, match it against an invoice or a contact you already trust.
A lookalike handle seems like an easy way to divert a payment. That's our inference. No source we read reports it happening.
3. Preview it, and a person confirms the send
Have the agent run the review step first, with --dry-run or the review output Stripe's skill describes, and paste the exact command and result into your chat. Check recipient, amount, currency and source account. Stripe's reference says: "Only add --yes after the user confirms the recipient, amount, currency, and source financial account." A person makes that call every time. The agent never adds --yes on its own.
4. Fund a separate small account for agent payments
Stripe Pay draws from a funded financial account. Give agent payments their own account and top it up on purpose. Your main operating balance stays out of reach, and the worst mistake is capped by what's in the small one.
5. Log every payment and review it weekly
For each send, record the handle, amount, fee and timing from the review step, who confirmed it, and the time. Add any send the agent proposed that a person turned down. Read the log once a week against your allowlist and caps.
Where this fits
Our money map note covers money moves on your store and your customers: refunds, cancels and purchases. This rule covers money going out to other businesses. If your agent reaches Stripe with a login you pasted, sort that out first with audit your agent's logins before your bank does.
What this doesn't tell you
Stripe hasn't published a newsroom post on Stripe Pay that we could find, so most launch details come through trade coverage. We don't know which countries have it, what the transfer limits are, or whether every route is free. The skill says to check the fee each time. Stripe hasn't said how transfers settle. We also couldn't confirm from the docs whether the same approval rules apply when an agent calls Stripe Pay from the CLI and from other tools.
What it does give you is a clear review step and a confirm flag, so you can put a person on every send from the start.
This week
Write the four line pay rule. Build the allowlist from invoices you already have. Open a small funded account for agent payments. Make the agent paste the review output before every send, and keep the confirm with a person. Start the log.
If you want your Always On Agents set up with pay rules like this from day one, get in touch.
Linked sources
- PYMNTS: Stripe Pay targets instant B2B payments (2026-10-27)
- RuntimeWire: Stripe Pay, profile based B2B transfers
- The Paypers: Stripe launches free instant B2B transfers via business handles (2026-10-28)
- Stripe Docs: stripe pay CLI reference
- Stripe: stripe-pay agent skill (stripe/ai, MIT)
- Related article: Draw your agent's money map before you connect it
- Related article: Audit your agent's logins before your bank does
Get our Articles and posts by email.
Up to one email a day with new Articles and posts from 22atlas.com and sean-weldon.com, plus occasional news about 22Atlas services.
Also on the desk
Technique
Draw your agent's money map before you connect it
Claude Duet launched with Shopify and Stripe on Oct 21, and the two handle refunds in opposite ways. Spend 30 minutes mapping every money action your agent can reach, then set the rules.
Technique
Audit your agent's logins in 20 minutes, before your bank does it for you
Banks have started blocking AI agents that sign in with the owner's password, and Chase now puts agent errors on the customer. List every login your agent holds and move each one to the right kind of access.
Technique
Write your store's agent access policy this week
BCG surveyed big retailers and found most have already picked a stance on outside shopping agents. Pick yours on purpose, draw the data line, and put it on one page.