Sierra, Meta publish personal agent protocol for AI commerce

On October 6, Sierra and Meta published the Personal Agent Protocol, an open standard intended to let personal AI agents authenticate with businesses, carry context across channels, and act on behalf of consumers. The announcement marks one of the most concrete industry moves yet toward a shared plumbing layer for agentic commerce.

The protocol is built on OAuth, the widely used authorization framework, and is designed so that a personal AI agent can prove who it represents, retain conversational and transactional context as it moves between channels, and operate through a company's website, an MCP or OpenAPI interface, or an agent the company itself operates. In practice, that means an assistant could start a support conversation in chat, continue it by voice, and complete a purchase without re-authenticating or losing the thread.

What the personal agent protocol does

At its core, the personal agent protocol standardizes the handshake between a consumer's agent and a business. Identity and consent travel with the interaction, so merchants can verify that an agent is genuinely acting for a specific customer and see the permissions that customer granted. Context portability means details established in one channel — an order number, a return reason, a delivery preference — follow the agent into the next one.

The specification also gives businesses flexibility in how they receive agent traffic. A retailer can expose its existing website, stand up an MCP or OpenAPI endpoint purpose-built for agents, or deploy its own company-owned agent that talks to personal agents on the customer's side. That multi-path design acknowledges that companies are at very different stages of readiness for machine customers.

Walmart, Shopify and Stripe anchor the founding coalition

The founding partner list reads like a cross-section of the commerce stack: Walmart, Shopify, Stripe, Rocket, Genesys and Instinct have all signed on, according to AI Weekly's October 7 coverage of the announcement. Their involvement signals that the protocol is being designed with real transaction volume in mind rather than as a purely theoretical exercise.

The effort also sharpens an emerging standards rivalry. Visa's Trusted Agent Protocol is pursuing a similar vision of authenticated agent commerce, and the two initiatives share some of the same partners — a sign that major players are hedging their bets while the field is still taking shape. Whichever approach gains traction, merchants appear unwilling to sit out the shift toward AI-mediated shopping.

The timing fits a broader wave of agentic retail experiments. TikTok has been testing an AI shopping assistant with one-click checkout, a move covered in depth on GenZ NewZ, while Constructor has added agentic checkout to its AI shopping agents, as previously reported. A shared authentication standard could stitch these experiments into a coherent ecosystem.

What comes next: a v0.1 spec, then payments and permissions

The initial release is only the starting point. A v0.1 specification is due later this month, giving developers and merchants a concrete document to build against. Payments and granular permissions are slated as future extensions — two of the hardest problems in agent commerce, since they determine what an agent is allowed to spend and precisely which actions it may take.

Those extensions will be closely watched. Granular permissions are what separate a genuinely trustworthy agent from a blunt automation: the difference between an assistant that can track a parcel and one that can issue refunds, change account details, or move money. Getting that consent model right is likely to decide how quickly consumers trust agents with high-stakes tasks.

Why it matters

For readers, the personal agent protocol is about what happens the next time a chatbot offers to handle something for you — rebook a flight, chase a refund, compare prices across stores. Today those handoffs are fragile; tomorrow, a shared standard could make them as routine as logging in with a social account. The companies now lining up behind the protocol are betting that the agent, not the app, becomes the primary interface between customers and businesses.