Meta and Sierra Develop Open Personal Agent Protocol for AI Agents

The problem: agents that click like people

Sierra says most personal agents today use websites and apps the way people do, by loading pages and clicking through forms. When that fails, they may call a company's support line or open its web chat. According to Sierra, this can take a long time and the agent may not finish the task. A direct connection, the company argues, could complete the same task securely in seconds.

Sierra describes three groups with different needs. Consumers want speed, dependability and trust. Brands want visibility and control, including knowing when a personal agent is acting for a customer and deciding what it can do. Companies building personal agents want efficiency and access through a direct, consistent way to work with participating companies.

How Personal Agent Protocol works

The guiding principle is that consumers decide what access their personal agents get, while companies set parameters for what those agents can do. The protocol is designed to handle authentication, empower consumers and give companies visibility into what personal agents do through their websites, APIs or company agents.

It starts on the company's website, where an agent can discover what the company offers and how to reach it. The agent then opens a session on the user's behalf. It can begin as a guest, which may be enough to check product availability or ask about a returns policy. If a task needs account access, the customer can sign in on the company's page or use credentials already set up with their personal agent. The customer decides whether the agent gets read-only or write access.

The session is built on OAuth, an established standard for authorizing access, and carries across channels. A question asked before sign-in and an order change made afterward count as part of the same visit.

From there, the agent can work through whichever route the company thinks gives the best experience:

• Its website: navigating the regular web pages.
• Its APIs: connecting through interfaces built on standards such as MCP and OpenAPI.
• Its agent: handling tasks that need conversation, such as a warranty claim.

Roadmap and partner views

Sierra says it plans to publish the v0.1 specification later this month, host design workshops with interested parties and release a reference implementation to help developers get started. Instinct is also joining the effort, and Sierra says it welcomes all partners.

Possible future additions include more detailed permissions that let customers and companies limit specific actions, push notifications that tell an agent the moment a flight is delayed or an order ships, and payments extensions that could let an agent complete a purchase without sharing credit card information.

Partners backed the effort in statements published by Sierra. Tony Bates, Chairman and CEO of Genesys, said brands need "a trusted way to know who an AI agent represents, what it's authorized to do, its intent, and how to work with it securely." Kevin Miller, Head of Payments at Stripe, said the company is contributing "to give businesses a standard way to recognize their customers' agents, efficiently interact with them, and shape their customer relationships." Mani Fazeli, VP Product at Shopify, said the protocol "helps establish how agents and merchants can work together."

What it means

As personal agents take on more everyday tasks, Sierra argues that companies need clear, secure ways to work with them. For consumers, the promise is faster task completion with control over what their agent can access. For businesses, it is visibility into agent activity and the choice of how to serve those agents. Whether the standard gains wide adoption will depend on the v0.1 specification and the participation of companies and agent builders.