Meta and Sierra, the customer-service agent company co-founded by former Salesforce co-CEO Bret Taylor, introduced the Personal Agent Protocol, a proposed standard for AI agents that browse and buy on behalf of consumers, SiliconANGLE reported Tuesday. Walmart and Stripe are among the partners, and Taylor said the specification is open for anyone to implement.

What Happened

Taylor announced the protocol in a post on Sierra's blog. According to SiliconANGLE, it covers how personal agents authenticate themselves to businesses, how consumers keep control over what their agents do and how companies can see what agents are doing on their websites and APIs. Taylor said the goal is to bring transparency to agent commerce so retailers know what a personal agent is doing on their sites, and he said he expects OpenAI to support the standard eventually.

David Singleton, a vice president at Meta Superintelligence Labs, likened the effort to email, which works because everyone uses the same shared standard to reach one another. TechCrunch, which also covered the launch, listed Genesys, Rocket, NiCE and Decagon as partners alongside Meta, Sierra, Walmart and Stripe.

RelatedEight States Sue OpenAI as Iowa Seeks Court Halt on Testing of Advanced AI Models

Neither report described a governance body or said when a full specification would be available.

Why It Matters

Consumer agents can only finish a purchase if the merchant lets them in, and many sites currently treat them like scraping bots. TechCrunch reported that agents including Meta's Muse, ChatGPT's Dots and Instinct have run into blocks on sites such as Amazon, Walmart, Delta, United, Yelp, eBay and Zillow.

The reasons vary. Amazon deliberately blocked Muse. Walmart told TechCrunch its blocks were an unintended result of human-verification checks and noted that it is a Muse partner. Yelp said it does not allow non-human traffic unless the agent has paid to license its data, and United pointed to terms of use that bar automated access without written permission. eBay said it restricts unauthorized agents and automated scraping rather than all shopping agents.

Infrastructure defaults play a role too. TechCrunch said a Cloudflare crawler-settings change on Sept. 15 led some sites that had been blocking AI bots to start blocking agents on pages that carry ads. Cloudflare declined to share data on how often personal agents are blocked.

Context

Pressure for common rules has also come from finance. Six large banks, including Bank of America and Capital One, recently published principles calling for agent standards built on transparency, safety, privacy, choice and interoperability, citing concerns about scams, fraud and agents that might favor commissions over users' interests, according to SiliconANGLE.

Meta has a direct stake in the outcome. Its Muse shopping agent, launched in September, needed emergency security fixes before release, has drawn privacy concerns over how it profiles friends and family and was banned by Amazon over identification problems, SiliconANGLE reported. It has also been a top App Store download.

The new protocol enters a field with other agent-commerce efforts. Stripe co-developed the Agentic Commerce Protocol with OpenAI in 2025, and Google introduced its Agent Payments Protocol the same year.

What to Watch

The test is adoption by the companies that control most agent traffic and online checkout. Watch whether OpenAI follows through on the endorsement Taylor anticipates, whether Google and Amazon join or keep their own approaches, and whether merchants that block agents today, including airlines and review sites, accept agents that identify themselves through the protocol.