What the Personal Agent Protocol means for a WooCommerce or headless store

The Personal Agent Protocol (PAP) is a proposed open standard for how a person's AI assistant signs in to a business and gets a task done there. Sierra announced it on October 6, 2026, developed with Meta, with Genesys, Instinct, Rocket, Shopify, Stripe and Walmart as partners. There is no specification yet: v0.1 is due later in October. If you run a store on Shopify, your platform is at the table. If you run WooCommerce, a headless storefront or a store behind your own CDN, nobody on the partner list serves you. This post covers what PAP is, what it isn't, and what an independent store should do this month.

What PAP is

Today, when someone asks Meta's Muse, Instinct or a similar assistant to "find me a quiet ceiling fan under $100 and order it", the agent uses your website the way a person does: it loads pages, clicks through forms and fills in a checkout. Some assistants phone or message the business instead. It's slow for the shopper, and from your side you can't tell an assistant acting for a real customer from a scraper.

PAP proposes a standard way in. Its organising principle, in Sierra's words: "Consumers decide what access to give their personal agents, and companies set parameters for what those agents can do."

A session, as Sierra describes it:

OAuth is the standard behind "Sign in with" buttons. The fifth step matters most for a store: you can keep agents on your existing pages, open an API to them, or put your own agent at the door for anything that needs a conversation. Sierra's example of the last is a warranty claim.

What PAP isn't

• Not a payment protocol. Sierra lists payments as a possible future extension. Until then, agent checkout runs through what exists today: checkout inside the assistant, or the agent operating your checkout page.
• Not finished. There's no specification or reference code yet; both are promised for later this month, with design workshops. Finer-grained permissions and push notifications are listed as future directions.
• Not universal. OpenAI and Anthropic aren't part of it, per CNBC, though Sierra co-founder Bret Taylor said he expects them to join. And not every business wants the agents in: Amazon blocked Meta's Muse agent from shopping on Amazon.com in September.
• Not the only standard. OpenAI and Stripe's Agentic Commerce Protocol puts checkout inside the assistant. Google's Universal Commerce Protocol, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, connects agents to stores' checkouts. Visa's Trusted Agent Protocol, developed with Cloudflare, helps merchants recognise trusted agents. On the same day as PAP, Decagon open-sourced PACT, a permissions protocol co-developed with Instinct, and joined the PAP working group. Stripe and Shopify sit at several of these tables. Expect overlap, and eventually a shake-out.

Who's at the table

Read the partner list as a merchant. On one side are the assistants: Meta's Muse and Instinct. On the other are businesses and the companies that serve them: Walmart, Shopify, Stripe, Rocket, and the customer-service platforms Sierra and Genesys. Decagon, another customer-service platform, has joined the working group.

Shopify is likely to handle PAP's discovery and sign-in for the stores on its platform, though it hasn't said how. Walmart will build its own. The customer-service platforms sell to large companies that run contact centres.

No one on that list sells to the WooCommerce store doing $3M a year, the headless Next.js storefront, or the brand on its own Cloudflare account. Unless someone builds it for them, those stores won't have a way to publish what agents may do, to decide which agents get guest or write access to what, or to keep a record of a PAP session when one arrives.

That's where independent stores already are with most things agent-related. Platforms serve their merchants, agent companies serve their agents, and the store in between sees the least.

What it changes for you

1. Your traffic gets a third category. Today you can separate, at best, declared crawlers from everything else. Once PAP sessions exist, some agents will announce themselves at the door and say which customer they act for. That could let you tie an order to a specific assistant acting for a signed-in customer, which matters the day that customer disputes it. It also sharpens the contrast with every agent that doesn't use PAP, which for now is most of them.

2. "Companies set parameters" needs somewhere to set them. The protocol says you decide. It doesn't give a small store the tools to decide which agents may open a guest session, which may get write access, on which products, during which windows (limited drops, say), and what happens to agents that don't identify themselves. Large retailers buy bot-management products for this. A smaller store has, at most, its CDN's bot settings, which see requests, not carts, products or orders.

3. Discovery becomes an asset. Step one of a PAP session is the agent finding out what you offer and how to connect. Sierra hasn't said what form that takes. Related standards use a file at a known address on your site: Shopify stores already serve a UCP profile at /.well-known/ucp and agent instructions at /llms.txt. If PAP works the same way, a Shopify store will get its file from the platform, and everyone else will publish their own.

What to do this month

Don't build against PAP yet. There's nothing to build against, and the spec may change between v0.1 and anything that ships in an assistant.

Do these four things. Each one helps whether PAP succeeds or not.

1. Measure agent traffic now. You can't set parameters for visitors you can't see. Know how much of your traffic is AI crawlers, agent browsers and people sent by AI answers, and which products they touch. Google Analytics excludes known bots, and you can't turn that off. Your analytics count an agent browser as a person.
2. Publish an llms.txt and check your product structured data. An llms.txt is the nearest thing to a discovery file today, and complete product data helps any agent that reads your pages.
3. Check what your product pages show without JavaScript. Many AI crawlers don't run it. If your price, variants or reviews only appear once scripts run, those agents are guessing.
4. Decide your policy on paper. Which agents would you welcome at checkout? Which only to browse? Are there products you'd never let an automated buyer take, like drops and limited runs? With the answers ready, you can act the day the tools exist, from your platform or from someone else.

What we're doing

Tactical measures AI crawlers, agent browsers and the visitors AI assistants send, on Shopify, WooCommerce and headless stores. On WooCommerce it also attributes orders to agents, with an evidence record for each one. That's step one above, available today.

The rules side, deciding what agents may do by product, cart and checkout, is next: drop protection and agent-checkout rules, coming.

When PAP v0.1 is published, we'll read it for what an independent store needs to do its part: publish what agents may do, decide which of them get guest or write access to what, and keep a record of each session. We'll build that for WooCommerce and headless stores, and we plan to ask to join the design workshops as an implementer for independent stores. Until there's a spec, nothing in Tactical works with PAP. We'll update this post when the spec lands.

Questions

Is PAP available now?

No. The v0.1 specification is due later in October 2026, with design workshops and a reference implementation.

Will my Shopify store get PAP?

Shopify is a partner, so it's likely the platform will handle discovery and sign-in for its stores. Shopify hasn't published details.

Does PAP handle payments?

Not in its first version. Sierra lists payments as a possible future extension.

Does PAP cover ChatGPT or Claude?

Not yet. OpenAI and Anthropic aren't part of it, per CNBC, though Bret Taylor said he expects them to join.

What should a WooCommerce store do about PAP today?

Measure agent traffic, publish an llms.txt, check what your product pages show without JavaScript, and decide your agent policy on paper. Don't build against an unpublished spec.