Meta and Sierra draft a sign-in standard so businesses can tell which agent is calling

On 6 October 2026 Meta and Sierra announced the Personal Agent Protocol (PAP), an open standard they are developing with partners Genesys, Instinct, Rocket, Shopify, Stripe and Walmart. It defines how a consumer’s AI agent authenticates with a business and what that business lets it do — the identity-and-permission primitive that agentic commerce has been missing. The v0.1 specification is due later in October; as of launch there is no published specification, licence or governing body.

What happened

PAP is designed around a simple split of control: consumers decide what access to give their agent, and companies set the parameters for what those agents can do. A session is built on OAuth, the same authorisation standard behind “sign in with” buttons, and carries across channels so context survives the moment a guest becomes a signed-in user.

Access comes in tiers:

  • Guest. The agent starts on the company’s website and can discover what the business offers — checking availability or asking about a returns policy without signing in.
  • Signed-in. For anything touching a customer’s account, the customer signs in on the company’s page (or uses credentials already set up with the agent) and decides whether the agent gets read-only or write access.

Once inside, the company chooses the route the agent takes: its regular website, its APIs via MCP or OpenAPI, or its own agent for tasks that need conversation, such as a warranty claim. Payments, push notifications and finer-grained permissions are listed as future extensions outside v0.1 — the first specification covers identity and coarse read/write scope only.

The backdrop is a governance fight. Meta launched its Muse consumer agent on 8 September; Amazon cut Muse off from Amazon.com on 20 September, saying Meta had not told it the agent would access the store and that the agent did not identify itself while browsing. “It is kind of chaos until such a standard exists,” Sierra co-founder (and OpenAI board chair) Bret Taylor said, comparing PAP to signing in with Google or Facebook credentials.

PAP also arrives into a crowded field: Visa’s Trusted Agent Protocol, Google’s Universal Commerce Protocol and OpenAI/Stripe’s Agentic Commerce Protocol already cover overlapping ground, and several PAP partners sit on rival efforts too. OpenAI, Anthropic, Amazon and Google are not on PAP’s partner list.

Why it matters for KYA

Know Your Agent is about establishing who an agent is, who stands behind it, and what it is authorised to do before it acts. PAP is a concrete attempt at the first two questions for consumer agents: it forces a personal agent to arrive as an agent, with a declared identity and a scoped, customer-approved grant, instead of a browser-in-a-cloud-VM typing a customer’s password into a login page. That directly addresses Amazon’s complaints about notice and self-identification.

Two caveats matter for anyone deciding whether to build on it:

  • Nothing is published yet. With no specification, licence or independent governing body, “open” is currently a stated intention rather than a verifiable property. A practical test is whether the v0.1 text ships a neutral governance model and at least one non-Meta agent implementation.
  • It stops short of the hardest part. Payments — the point where agent authority becomes financially irreversible — are explicitly deferred. Until then, PAP answers “which agent is this and may it read or write?” but not “may it pay, and how much?”.

For a business, the useful work does not depend on the winner: decide what an unauthenticated agent may see, what a signed-in agent may read, and what it may change, and map those tiers onto OAuth scopes. That matrix ports to PAP, UCP or TAP with little rework.

Sources


Know Your Agent (KYA) explains agent identity, verification and accountability. This is an explainer, not legal or compliance advice — see our Sources & methodology. New to KYA? Start with What is Know Your Agent? and the glossary.