Overview
Personal Agent Protocol is the open standard Sierra and Meta announced on October 6, 2026 with industry partners at Genesys, Instinct, Rocket, Shopify, Stripe and Walmart. It addresses something that has become urgent rather quickly: as people start sending their own AI agents to book appointments, reserve flights and shop for car insurance, how does a business receive a visitor that is not a person but acts for one.
Today most personal agents use websites and apps the way people do, loading pages and clicking through forms, and when that fails they may call the support line or open a web chat, which is slow and often does not finish the job. This protocol offers a direct connection instead, letting an agent work within what the business permits while leaving control in both sets of hands.
The principle is that consumers decide what access to give their agents, and companies set the parameters for what those agents can do. It is open for anyone to implement.
Key Features
- Discovery starts on the website: An agent first discovers on the company website what is offered and how to reach it, then opens a session on the user's behalf. The business does not need a separate entry point, since the existing site is the discovery layer
- Sessions built on OAuth: The session rests on OAuth and carries across channels, so a question asked before sign-in and an order change made afterward belong to the same visit. When a task needs account access, the user can sign in on the company page or use credentials already set up with the personal agent
- Start as a guest, escalate on demand: It can begin as a guest, which is enough for checking stock or asking about a returns policy. When account access is required the user decides between read-only and write, keeping the switch in human hands
- Three routes to get the job done: The company chooses the path that fits: the agent navigates the regular web pages, connects through APIs built on standards such as MCP and OpenAPI, or works through the company's own agent for tasks that need conversation, such as a warranty claim
- Built for three sets of needs: Consumers want the job done right the first time by an agent they can trust. Brands want to know when an agent is acting for a customer and to decide what it may do. Agent builders want a direct, consistent way to reach participating companies
- Open, with adopters already at the table: Open for anyone to implement, with Genesys in customer service, Shopify in commerce, Stripe in payments, Walmart in retail, and Rocket and Redfin in homeownership among the participants. The v0.1 specification and a reference implementation are planned for later this month
Use Cases
- Retail and commerce: let a customer's agent check stock, place orders and process returns without simulating clicks through a storefront
- Contact centers: recognize that an agent is acting for a real customer, authorize it within preset bounds and keep full visibility
- Long-running journeys such as property and financial services, where an agent moves through several stages in one session, from finding a home to securing financing
- Agent developers: integrate with many businesses through one protocol instead of writing a bespoke adapter for each
Pros
- Led by Sierra and Meta with participants spanning support, commerce, payments, retail and property, so adoption looks more real than a paper standard
- Built on OAuth rather than a new mechanism, so existing authorization infrastructure connects directly
- Three integration routes let companies take part on their own terms instead of rebuilding everything at once
- Permissions are set by the user and split between read and write, while brands keep visibility and control
- An open standard anyone can implement, with no single company owning the entry point
- A published timeline for the v0.1 specification and reference implementation means developers can start now
Pricing
Offered free as an open standard with no licensing fee, and the specification and reference implementation can be implemented freely once published. Cost comes from the engineering work on the company side to expose interfaces, wire up OAuth and accommodate agents, plus the resources to run those interfaces. The v0.1 specification and reference implementation are planned for later this month.
Summary
The weight of this protocol lies less in its technical difficulty than in who is sitting at the table. Genesys in support, Shopify in commerce, Stripe in payments, Walmart in retail, Rocket and Redfin in homeownership, plus Sierra and Meta, cover most of what people actually send personal agents to do: ask, buy, return, claim, complete paperwork. Whether a standard gets used depends on whether it satisfies all three parties at once, and this list suggests the intent is there.
Technically it takes the least risky path available. OAuth already exists, website discovery already exists, MCP and OpenAPI already exist, and the protocol assembles them into a route everyone recognizes rather than inventing new machinery. That lowers the barrier for businesses, and it also means the real work lands in designing backend permissions and visibility rather than in the protocol itself.
What is worth watching is the stated next step. Finer-grained permissions would let customers and companies set limits on specific actions, push notifications would let a company tell an agent the moment a flight is delayed or an order ships, and payments extensions would let an agent complete a purchase without sharing card details. Any one of those landing would visibly widen what an agent can do.
Right now this is an announced protocol with the v0.1 specification and reference implementation still on the way. If you build personal agents, this is the moment to read the specification and join the design workshops, because adapting after a standard like this settles costs far more.
Version History
- Personal Agent Protocol announced (2026-10-06): Sierra and Meta announced Personal Agent Protocol with partners including Genesys, Instinct, Rocket, Shopify, Stripe and Walmart: an open standard defining how personal AI agents interact with businesses, open for anyone to implement. An agent discovers on the company website what is offered and how to reach it, opens a session on the user's behalf over OAuth, starts as a guest, and lets the user decide between read-only and write access once an account is involved, then completes tasks through the company website, APIs built on standards such as MCP and OpenAPI, or the company's own agent. The v0.1 specification, design workshops and a reference implementation are planned for later this month, with finer-grained permissions, push notifications and payments extensions named as next steps