AI agent chaos: Sierra and Meta propose a common standard, but fragmentation is already here

On October 6, 2026, Sierra's co-founders announced the Personal Agent Protocol (PAP) on the company blog. Developed together with partners such as Meta, Genesys, Shopify and Walmart (although the list varies slightly between sources), the protocol aims to resolve a growing tension: the one between personal agents that browse and shop online on users' behalf and the corporate platforms trying to protect their own boundaries. According to Implicator.ai, on September 20 Amazon blocked Muse, Meta's personal agent, from accessing Amazon.com, arguing that it had not been informed and that the agent did not identify itself while browsing. Meta, for its part, says credentials are kept in a secure vault. As Sierra co-founder Bret Taylor told CNBC, according to Implicator.ai and Forkast, until there is a common standard the situation will remain «kind of chaos».

The protocol sets out to govern authentication, user control and the way a company exposes its services. Built on the OAuth standard, a session would let a digital assistant start browsing as a guest and then log in, leaving the user to decide whether to grant read-only or also write permissions. At launch, however, the technical details are still on paper: version 0.1 of the specification is expected by the end of October 2026, to be followed by design workshops and a reference implementation. Neither the license nor the project's governance has been clarified yet. Moreover, Meta's participation is documented by Sierra's post and by statements to the press, while the social media giant has made no dedicated official announcement of its own.

The bet placed by Sierra and its partners, though, has to contend with an already fragmented market. According to Implicator.ai and Forkast, OpenAI, Anthropic and Amazon are not among the partners, and it is unknown whether they will join. For Genesys's CEO, personal AI is creating «a new front door to the enterprise». At the same time, companies like Shopify and Stripe are already part of competing initiatives such as Visa's Trusted Agent Protocol, while Walmart and Shopify itself back Google's Universal Commerce Protocol. More than a universal agreement, the landscape looks like a race to decide who will control the rails future shoppers will travel on.

Regulating the behavior of AI agents through authorized, transparent channels is a sensible step if we want to keep the web from turning into a clash between bots and corporate barriers. But the real challenge is not technological, it is adoption: an open standard only has value if everyone agrees to use it. As long as several competing protocols exist, pushed among others by Visa, Google and OpenAI with Stripe, this front door risks remaining a side entrance. — Olya