x401 Explained: Proof’s AI Agent Identity Protocol

Last updated: October 15, 2026

Quick answer: Proof's x401, launched 15 October 2026, is an open protocol that verifies who an AI agent acts for and signs what the user approved. Here's how it works.

On 15 October 2026, an identity company called Proof put a candidate answer to the agent-identity fight into production: x401, an open protocol that lets any website ask an AI agent to prove who it is acting for — and prove that the person approved exactly what the agent is doing. The timing is pointed. It comes weeks after Amazon blocked Meta’s Muse agent from its marketplace, exposing the problem Proof calls the “agent evidentiary gap”: an agent may have a valid password, token or card on file, but none of those proves the person behind it authorised this particular action. This piece explains what x401 does, how it works, and where it fits beside the other attempts at the same problem — including Meta and Sierra’s agent sign-in efforts. It is the follow-up to our Amazon–Muse explainer: the fix, or at least a proposal for one.

What x401 actually does

Today, when an AI agent hits a website, the site sees browser traffic that looks like a person. It can check a password or an OAuth token, but it cannot check intent. x401 borrows the mechanism of HTTP 401 — the “unauthorised” response the web has used for decades — and extends it: a website answers the agent with a challenge that names the evidence it must provide before proceeding. That evidence can be a verified identity claim (legal name, age, country of residence) or a cryptographic signature binding a specific person to the specific terms of a transaction.

Three things went live with the launch, according to Proof’s announcement: Proof’s agent identity wallet (the app where a person sees what’s being requested and approves it on their own device), the x401 protocol itself (the open HTTP scheme for requesting and verifying that evidence), and a production MCP server that works with AI assistants that support the Model Context Protocol — including Meta’s Muse, ChatGPT dots, Grok and Claude.

How x401 works: request, approve, verify

The flow, as described in Proof’s protocol documents, has three steps:

  1. Request. The website challenges the agent: to proceed, it must present a verified identity claim at a required level of assurance, or a signed copy of the transaction’s terms — for example, the amount, recipient and source account of a wire transfer.
  2. Approve. The agent shows that request to the person on their own device. The person authenticates — Proof says with a biometric — and sees exactly what is being asked: the identity attributes and the transaction terms. On approval, a hardware-protected key signs the presentation, binding the person’s verified legal identity to the terms they approved.
  3. Verify. The website checks the issuer, the credential, the signature and the transaction terms before executing anything. Because the evidence is signed, the site can keep it and re-validate it later if the transaction is disputed. If the agent altered the amount or the recipient after signing, the signature no longer validates.

The important design decision: this does not just establish who someone is. It creates tamper-evident evidence of what they agreed to. Proof illustrates it with a $25,000 wire: the bank’s x401 challenge requires the signature on the wire instructions themselves, and the bank verifies that signature alongside its normal fraud checks before the money moves.

Why the launch leans on law, not just cryptography

Crypto alone doesn’t make an agent’s signature legally meaningful. Proof’s announcement spends real effort on the legal layer, and three of its claims are worth understanding:

Proof also built x401 to inter-operate with the identity-and-payment neighbours rather than compete with them: x402 (the HTTP-native payment protocol now stewarded by the Linux Foundation, with more than 40 member organisations including Visa, Mastercard, Stripe, Google, AWS, Circle and Coinbase as of July 2026) carries the payment, while x401 carries identity and signed authorisation. Basic identity claims can be requested without a sales process, and websites can pay Proof per claim over x402 — identity metered like bandwidth, one verification at a time. x401 is also designed to interoperate with AP2 and Verifiable Intent, the specifications from Google and Mastercard now stewarded by the FIDO Alliance.

x401 vs the other answers: who’s trying to fix agent identity

x401 is not the first attempt. Here is how the serious candidates compare — the table is our own synthesis; the facts come from the announcements and reporting linked below.

Approach What it verifies Who runs it Status
x401 (Proof) Verified legal identity (IAL2) + cryptographic signature on the exact transaction terms Open protocol; Proof issues credentials Live 15 Oct 2026; Superfluid first live implementation (Sept 2026)
Signed agent requests (Cequence “Agent Trust”) Whether an agent presents a verifiable identity vs. traffic that looks human Security vendors observing traffic Live; Cequence published Muse-traffic findings 8 Oct 2026
Personal Agent Protocol / “Poppy” (Meta & Sierra) Agent sign-in standard — agents identifying themselves to sites Industry standard effort In development
Passwords, OAuth tokens, saved cards That someone has access credentials — not intent Everywhere The status quo that created the gap

The honest way to read this table: x401 is the most legally-grounded answer yet, and the only one in production that ties a real-world identity to the signed terms of a specific action. It is also the most demanding of the user — every consequential action means a biometric approval on your own device. Whether people and merchants accept that friction is the open question.

The three-question trust test (our take)

Regardless of which protocol wins, this launch gives you a practical way to judge any system that lets an AI agent act for you. Before you hand an agent your money or your identity, ask three questions:

  1. Who can prove I’m the one behind this agent? A password proves access. Only a verified identity bound to you — the way x401’s certificate does — proves it’s you. If a service can’t tell the difference, its fraud model has a hole.
  2. What exactly did I approve? “I authorised a purchase” is not the same as “I authorised this item, this amount, this recipient.” Signed terms — where changing any detail invalidates the approval — are the standard worth demanding.
  3. Who can check later? Disputes happen. Evidence that the merchant can retain and independently re-validate (x401’s stored signed evidence) beats an audit log only one company controls.

Our earlier agent-safety checklist covers what to do before an agent runs anything; these three questions cover what to demand of the platforms once money is involved.

What we couldn’t verify

Three honest caveats before anyone treats this as settled. First, everything above comes from Proof’s own announcement — the GlobeNewswire release reproduced across news wires on 15 October 2026. It is a vendor launch document, not an independent review; no third party has audited x401 in production yet, and Proof has not published independent adoption numbers. Second, availability is unannounced beyond the US-centric framing: identity proofing starts from government-issued ID, and Proof has not stated which countries it supports. Third, the economics are unknown: per-claim pricing over x402 is positioned as the business model, but no prices are public, and no major merchant or bank has announced x401 support. As Proof’s own CEO put it, “blocking agents is understandable, but it’s not sustainable” — the industry agrees on the problem; the protocol that sticks is still undecided. We will update this piece when independent implementations or merchant adoption are announced.

Frequently asked questions

Do I need Proof’s wallet to use x401?

Consumers connect Proof to their AI assistant and enrol for a credential when one is first requested, per the announcement. The wallet is where you see and approve requests; the credential belongs to you, not to an institution.

Does x401 work with Meta’s Muse?

Proof lists Meta’s Muse, ChatGPT dots, Grok and Claude among the assistants its production MCP server supports. But note: this does not resolve Amazon’s block of Muse — Amazon’s dispute is about its own terms of service, not about whether Muse can present an identity. A protocol only helps if the merchant accepts it.

Is x401 a payment protocol?

No. x401 establishes who is behind a request and what they approved; payment travels over x402, the separate HTTP-native payment protocol stewarded by the Linux Foundation. The two are designed to work together.

Is x401 better than just letting my agent log in with my credentials?

For consequential actions, yes, in principle: a saved credential proves access, while x401’s signed terms prove you approved this specific action. That said, per-claim verification may carry costs, and everyday low-stakes tasks don’t need legal-grade identity.

Sources and methodology

Sources: Proof launches x401, giving AI agents a way to prove who they represent — and what they’re authorized to do (GlobeNewswire, 15 Oct 2026) — the launch announcement this article is based on; Manila Times reproduction of the release (15 Oct 2026). Related reading on this site: Why Amazon Blocked Muse: The Agent Identity Fight Explained, AI Agent Safety: 7 Checks Before It Goes Rogue, Meta Muse Privacy Explained.

Method: claims rest on Proof’s 15 October 2026 launch announcement and the regulatory guidance it cites; vendor claims are labelled as such. Corroborating independent reporting on adoption, pricing and international availability was not found at publication time. Verification date: 15 October 2026. Desk-researched; not hands-on tested. This is an explainer, not financial or legal advice.

Have a burning question about this topic?
Feel free to email us at contact@openaimaster.ai — we are happy to help!