A portable, merchant-owned reliability record that AI shopping agents read before they buy.
Soon your assistant does the buying, not you. When an agent has a thousand merchants for one product, a single question decides the sale: which merchant does it trust?
The checks that exist are gameable. The one that matters has no answer.
"None of the protocols address merchant trustworthiness, reliability, or likelihood of fulfilment." - confirmed across all six, from their own specs.
As real money moves through agents, today's signals become the attack surface - and the defence isn't another gameable badge, it's reliability measured from real outcomes, read from a neutral source no merchant can spoof.
Across every protocol, merchant reliability - will it ship, refund, not defraud - is out of scope. Even the most advanced independent effort verifies merchant identity and explicitly defers the reliability score.
They verify who the merchant is. Nobody measures whether it delivers.
No agent trusts a rival's say-so. The only trust layer every agent can read is the one no platform owns. UCP is Google-controlled and donated to no neutral body - the centralized seat is taken; the neutral one is wide open.
Neutrality is a position the giants structurally cannot occupy.
A physical store, a Google Business Profile, and multiple 5-star reviews. Visible, trusted, choosable.
Fulfilment, no retail - so Google revoked the Business Profile, and every review vanished with it. A reliable store, invisible to agents - for a reason that has nothing to do with whether it delivers.
Trust today is rented from a platform and revocable at its whim. We're building the reputation a merchant owns and carries - because we're the merchant who lost it.
Built from real sales, deliveries, refunds and disputes over time - measured from what actually happened, not declared by the merchant. Hard to game: the facts are attested by independent sources - you can't fake what Stripe signs. Graded, so an agent can choose between merchants.
A signed statement of fact. "Stripe attests: 12,400 orders, 0.3% dispute rate." Vouched for by a party with no reason to lie.
The tamper-proof envelope. Issuer signs, merchant holds, any agent verifies - without phoning the issuer.
A self-owned identifier - a W3C web standard, not a token or a chain. No platform issues it or can revoke it; it's the mechanism behind "the merchant owns it."
The merchant connects once. The record refreshes itself, travels everywhere, and any agent can verify it in milliseconds.
Published as a W3C Verifiable Credential - the same envelope Google's AP2 and Mastercard already use - at the merchant's own domain and in its UCP manifest. UCP agents read it through UCP; everyone else reads it directly.
Cryptographically signed, so nothing on the merchant's page can override it. UCP-compatible, never UCP-dependent.
Watch public signals over time - price/stock consistency, review authenticity, store age, policy changes. No permission needed. A baseline record on millions of merchants from day one.
The merchant connects real Shopify, Stripe and courier data to upgrade to a measured record.
Thin file to full file - exactly how a credit bureau already works. The bureau is populated before anyone opts in.
The intersection - measured, portable, neutral merchant reliability - is the empty seat. That's us.
A $1T+ channel by 2030. The comps (S&P, Moody's, Experian, D&B) prove trust-data is durable and high-margin - the shape, not a number we inherit. Our edge isn't their regulatory mandate; it's first-mover data accretion, neutrality, and being agent-native.
The record is free to read at the point of purchase, forever. Meter the lookup and agents route around it - ubiquity is the moat.
Pay to be measured, verified and equipped - distribution, conversion, tooling. Never to move the score. Trusted = chosen = more sales.
Platforms, processors and underwriters license the dataset (SLA feed, not per-query). They earn on trust decisions at scale.
Revenue never depends on charging for a validation - the one thing that kills adoption. Underwriting transactions with the data is later optionality, not the plan; the raise rests on merchant subscriptions, not a someday-fintech.
Illustrative ARR. The Y4-Y5 jump is the data-access layer compounding on the merchant base as agent volume arrives - not headcount.
The wedge pays the bills while the forcing function matures. We don't wait for the market - we sell into the one that exists now.
A measured-reliability dataset that deepens with every merchant and every month. It can't be back-filled or copied.
The Switzerland position no platform can occupy - rivals can't be each other's trust authority.
The record is merchant-owned and portable - they carry it everywhere, which is why they adopt and stay.
First mover accretes the data and becomes the default reference before any incumbent can turn.
We publish measured, independently-attested facts (ship, refund, dispute rates) - never a "safe / unsafe" judgment. Sourced facts are defensible; verdicts get you sued.
Every merchant sees and can challenge its own record - the credit-bureau correction process. The mechanism that limits liability is the same one that earns neutrality.
We're the data, not a guarantor. Methodology is public; the agent makes the call. Designed in from day one, not bolted on.
The one thing a trust business gets sued over, turned into a credibility and neutrality feature - with the regulatory homework (FCRA-style governance) done early, not discovered late.
Merchant-zero is KAAL - the same store Google stripped of its reviews. We gave it back a reputation it owns, measured and signed, and asked the agents.
On KAAL's real data, the v1 loop runs end to end: measurement engine → signed reliability credential → published agent-readable, plus the observe-engine and the test harness.
Pre-registered, 3 models. Handed the signal: KAAL 0%→92% (vs 3% for the same numbers self-declared). Given just a fetch tool, no nudge: agents fetched it unprompted 100% of the time, picked KAAL 70%, dropped to 0% on bad numbers. They read the content, not the label.
Power up the forgery test (clean early pass in hand) to deck-grade n, a paying non-KAAL merchant, then live surfaces.
Honest scope: LLM-agent proxies in a sandbox, not live shopping surfaces. A first in-loop-verify run is a clean early pass - agents took the genuine credential and rejected both the forged and the self-declared one (3/3 vs 0/3; small n, powering up).
Team will grow; these are the key players. Commerce, identity & attestations, and cryptographic engineering - with two prior exits between them.
We've now shown agents use a verified reliability signal when they see it - 0 to 92%. So the question is no longer "will it work." It's timing and reach: when agents fetch it unprompted and weigh trust at scale in the wild. We don't bet the company on that clock - the wedge pays on today's demand while we accrete the data moat.
The mechanism is proven. The remaining bet is how fast the world arrives - and the wedge pays while it does.
18-24 months to: v1 proof → 100-500 design partners → proven CAC/LTV → first independent issuer → Series A.
The trust layer is inevitable. It can't be the platforms. It should be owned by the merchants and built on real attestations.
Let's build it.