Autonomous agents are an entirely new category of users to build for, and, increasingly, to sell to.
Today, we’re launching (a preview) of machine payments on @stripe—a way for developers to directly charge agents, with a few lines of code. 🤖💸
$ Let’s start tinkering… ⤵️
You may be wondering: “Why can’t agents use virtual cards and transact as people do?”
Agents need:
- microtransactions
- 24/7 global rails
- controls (for human out of the loop)
- http native
- low latency
- finality guarantees
The current financial system is tuned for humans.
Now, businesses can accept machine payments via our regular ol' @stripe PaymentIntents API.
Charge agents for their API usage, MCP calls, or HTTP requests with agent-specific pricing plans, alongside your conventional subscriptions and invoices. Here's a demo from @stevekaliski:
The flow:
1/ Create a PaymentIntent 2/ Stripe generates a unique deposit address per transaction 3/ Return the address to an agent and instruct it to send funds or a payment token 4/ Track transaction status via API, webhook, or Dashboard 5/ Funds settle in your default balance
We’re launching with support for x402 using USDC stablecoins on @base, with more protocols, payment methods, currencies, and chains to come. Let us know which to prioritize.
And sales tax, refunds, and reporting just work. (You only need to think about crypto if you want to!)
We’re also releasing an open source cli called `purl` for you (and your bots) to test machine payments in the terminal, along with Node and Python samples. (Yes, “payments” + “curl” creatively smushed together.)
Stripe is building economic infrastructure to help businesses thrive in the soon-future world where agents, humans, and business all transact. There’s billions of us but we expect trillions of agents (some now clawing on Mac minis).
The TAM of the internet is about to increase.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
You can now send, sign, and track SAFEs in a few clicks, right from the @atlas dashboard.
Sincerely,
The Stripe team
Business money can only come from three places:
- cash from operations (revenue - costs)
- debit financed from loans
- equity fundraised from investors
In addition to our revenue (Stripe Payments) and loan (Stripe Capital) products, we want to help founders raise (via SAFEs).
We have major plans for making fundraising easier, faster, and hopefully more fun—globally—and would love your feedback and feature suggestions. Big and small ones, please.
Hi! We have three big agentic commerce announcements to share. @stripe is:
1/ powering Instant Checkout in ChatGPT 2/ releasing an open standard: Agentic Commerce Protocol, codeveloped with @OpenAI 3/ shipping an API for agentic payments: Shared Payment Tokens
Let's dive in ⤵️
Purchase where you prompt. Millions of people can now shop through ChatGPT, powered by @stripe.
- It’s fun to discover products via LLMs
- You can buy directly in chat from @etsy, soon @shopify, then more
- Use a saved payment method (I like @link!) or new one
- Buy in one tap
It's built on the Agentic Commerce Protocol (ACP) by @stripe and @OpenAI
- An open standard for programmatic flows between buyers, AI agents, and businesses
- Make your checkout agent-ready
- Works with any agent, commerce platform, payment processor
Today, @OpenAI launched their Agents SDK developer framework.
Until now, agents have mostly been chat-oriented (chat in, chat out), but agents will increasingly be action-oriented (data in, action out).
We, @stripe, would love to show you a few financial agents we’ve built! ⤵️
But, first, what is OpenAI Agents SDK?
It is now easy to build *multi*-agent workflows, with some nice features:
- Handoffs: Agents call each other and pass along context
- Guardrails: Validate inputs/outputs
- Tools: Files, Web Search, and Computer Use (same tech as Operator)
Last November, eons ago in AI-time, we launched our agent toolkit to help LLMs use @stripe. It is now being downloaded thousands of times per week.
Today, we've updated our toolkit to work with OpenAI’s framework.
📺 Here’s a few Stripe agents we've built (with sample code):
1/ Agents and LLMs can now understand, query, and execute @stripe, with our new MCP server.
Boot it up with a single, copy-and-pasteable line in any tool, like your AI IDE or client. Let's hop in! ⤵️
2/ Okay, MCP, what's the big deal?
LLMs know general, though sometimes out-of-date stuff (thanks pre-training!), but you often want to:
- understand latest, official info ()
- query your own data (your customers)
- execute API calls (create products)docs.stripe.com
3/ Along comes MCP (Model Context Protocol), open source from @AnthropicAI.
MCP is a standard way for any application to connect to an LLM — "a USB-C port for AI apps". LLMs now can *find* and *use* third party APIs and services.
You start a company and your founder friends yell: “OMG FILE YOUR 83B!”
Why?
Imagine your startup grows to $100M as your shares vest (wahoo!) and you're now in the top income bracket—you could save *several million dollars* in personal income taxes.
How does an 83(b) work?
The US wants to encourage entrepreneurship, so the IRS lets you "elect" to pay tax on your unvested shares upfront (when the value is likely low), so then:
- You won't owe any income tax as shares vest
- You only owe capital gains tax later if you sell