Your Customers’ AI Agents Are Coming. Meta and Sierra Just Proposed the Rules for How They Sign In.

On Tuesday, Meta and Sierra announced the Personal Agent Protocol, a proposed open standard for how a customer’s personal AI agent signs in to a business and what it is allowed to do once it gets there. Walmart, Shopify, Stripe, Genesys, Rocket, and Instinct are listed as partners. It lands about two weeks after Amazon blocked Meta’s Muse agent from its marketplace, which we covered on September 24. That story showed what happens when a platform shuts agents out. This one is the industry trying to build the front door instead. If customers start sending agents to research, book, and buy for them, the question for your business is whether you are ready to be read by one.

What Was Announced, and What Was Not

This is an announcement, not a finished standard. The v0.1 specification is due later in October, along with design workshops and a reference implementation for developers. Payments are not part of v0.1. They are planned as a later extension, along with proactive notifications such as flight delays or shipped orders.

It also has a competitor. Visa’s Trusted Agent Protocol, launched in June, reportedly lets agents buy from about 175 million merchants, and Stripe and Shopify are in both camps. Two overlapping standards from major players is a good reason to watch before you build anything.

How It Would Work

The protocol is built on OAuth, the same sign-in standard behind most API connections today. An agent could start as a guest, enough to check stock or the status of a return, then move to read-only or write access after the customer signs in and approves it. The customer decides how much access the agent gets.

The business decides which actions an agent can take and how it connects. That could be through its regular website, through its APIs using standards such as MCP and OpenAPI, or through a business-run agent that handles a task like a warranty claim. Sierra co-founder Bret Taylor has said the goal is to give businesses visibility into when an agent is acting for a real customer, and to help tell those agents apart from unauthorized bots.

Why This Matters Before It Is Finished

Sierra says most personal agents today work the way a person does, loading pages and clicking through forms. That is already happening on your website, whether or not any standard exists. A site with clear hours, pricing, availability, and policies is easier for an agent to read correctly. A messy one is easier for an agent to get wrong. The protocol will make this more formal, but the underlying issue is here now.

What to Do Now

Do not rebuild anything this week. A spec that has not been released and a competing standard are not a foundation to build on. Three cheap steps are worth taking instead.

Read your own website as if you were an agent. Check that the basics are accurate and easy to find, including services, pricing or pricing ranges, hours, how to book, and how to reach a human.

Decide in advance what you would let an agent do for a customer, such as checking an order or booking a consult, and what should always involve a person, such as pricing a custom job or handling a complaint.

Keep a human path open. Customers will still want one when an agent gets stuck.

The Honest Takeaway

This is early, and standards like this can stall or get replaced. But the direction is clear enough to plan around. The next customer to find you may arrive through software acting on their behalf, and that software will treat your site like a database before it treats it like a brand. Businesses that keep their information accurate and easy to act on give those agents less room to get them wrong. That is a marketing problem as much as a technical one.

---

If you want the background on why platforms are fighting over agent access, this connects directly: Amazon Just Blocked Meta’s AI Shopping Assistant. Here’s the Platform Risk Every Founder Should Understand.