Enterprise Tech

Meta and Stripe Back Personal Agent Protocol for AI Agents

Meta, Sierra, Stripe and Walmart are developing Personal Agent Protocol to define how AI agents access business services, with a first specification due in October.

By Daniel Mercer Edited by Maria Konash Published: Updated:
Meta and Stripe Back Personal Agent Protocol for AI Agents
Meta, Sierra and Stripe are developing shared rules for personal AI agents interacting with businesses. Photo: Julio Lopez / Pexels

Key Notes

  • Meta and Sierra are developing Personal Agent Protocol with Stripe, Walmart, Shopify and other partners.
  • The proposed OAuth-based sessions would connect customer consent with business-controlled access.
  • Version 0.1 is due later in October, while payment extensions remain a future possibility.

Meta, Sierra, Stripe, Walmart and Shopify are backing Personal Agent Protocol, a proposed open standard for personal AI agents dealing with businesses. The initiative aims to give companies a consistent way to recognize a customer’s agent and understand what that customer has authorized it to do.

Stripe confirmed its participation alongside Genesys, Instinct and Rocket. Meta and Sierra are developing the protocol, with an initial specification planned for later in October.

The practical issue is familiar to anyone delegating a task: permission to investigate a purchase does not necessarily include permission to complete it. As software takes on bookings, orders and support requests, businesses need a way to distinguish those instructions and connect them to the right customer.

What Personal Agent Protocol Would Do

Sierra’s October 6 announcement describes customer-controlled read-only or write access. A session could begin as a guest, with account authorization added when the task requires it. The business would retain control over which capabilities it makes available.

For example, a customer might allow an agent to check an existing reservation while withholding permission to change it. A separate instruction could authorize a modification. That is an illustration of why access boundaries matter; the detailed rules for individual actions still need to be specified.

The proposed session could continue across channels, connecting an initial question with a subsequent account action. Businesses could serve agents through their websites, APIs using standards such as MCP and OpenAPI, or conversational agents of their own.

Why OAuth Is the Foundation

The session would be built on OAuth, the established framework for delegated access. In the OAuth specification, access tokens represent authorization with particular scopes and durations. The relevant servers enforce those limits when an application requests protected information or an account action.

In plain language, a service can give software permission to perform a defined set of tasks without treating every request as unrestricted account access. A scope describes the access being granted. The authorization server can limit that access according to its policies and the account holder’s instructions.

For personal agents, that provides a useful foundation for expressing a customer’s consent. It does not establish that the agent has understood the customer correctly or chosen a sensible action. A valid authorization and a good decision are separate requirements, which businesses will have to evaluate in their own workflows.

Why Retailers and Agent Developers Want Shared Rules

Meta’s David Singleton, in a public post about the initiative, compared the opportunity with email: common rules can let different systems communicate without a single company owning the whole exchange. He framed the protocol as a way to improve experiences for people using Muse and the businesses serving them.

That helps explain why the partner list brings together personal-agent developers, retailers, commerce platforms and customer-service providers. A shopping task can touch several systems before it is complete. Shared conventions could reduce the amount of special integration work needed each time an agent reaches a new business.

Our coverage of Muse shopping examined the distinction between finding a product, preparing checkout and approving a purchase. Personal Agent Protocol addresses the broader relationship around those tasks: how the business recognizes the delegated customer and what access it accepts.

A useful implementation would also need to handle an ordinary failure clearly. If a requested booking change is unavailable, the agent should receive an understandable response and know whether it can try an alternative. Those practical outcomes will matter more to customers than whether the interaction happens through a webpage or an API.

The First Specification Is Due in October

Sierra plans to publish version 0.1 later this month, followed by design workshops and a reference implementation. More detailed permissions, push notifications and payment extensions are described as possible future additions. Those additions remain part of the protocol’s development roadmap.

The payment proposal could eventually allow purchases without sharing credit card information directly. For now, the immediate milestone is the specification itself: developers need its precise requirements before they can judge interoperability, implement access controls or assess how it fits their existing systems.

The commercial test will be whether participating services can complete real customer tasks reliably while respecting the permissions granted. A broad set of backers gives the project a starting point. Adoption will depend on the published design and the experience businesses and customers get when those rules are put to work.

Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.

AI & Machine Learning, Enterprise Tech, News