The acceptance layer for agentic e-commerce.

An AI agent arrives at an online store with a loaded wallet and an instruction to buy. The store is open. The catalog is published, the prices are current, the checkout works. And nothing happens, because the two sides have no handshake. The store cannot tell the agent apart from a scraper, cannot verify who it is acting for, and has no machine-readable way to quote it, reserve inventory for it, or accept its payment.

This is not a demand problem. The demand is already at the door and growing. It is an acceptance problem, and it is the reason a merchant sees rising automated traffic and falling attributable revenue at the same time.

Every component needed to connect an agent to a checkout was standardized inside the same eighteen months, and each one solves exactly one layer. None of them composes with the others on its own, and no merchant is going to assemble them.

The raw standards exist and are open. Composing them into a single thing a business can switch on is the opening, and it is open right now.

Facet terminates all of it on the merchant's behalf. The agent side speaks the open protocols it already speaks: a verified agent identity at the door, an RFC 9421 signed request, an MCP or UCP checkout, settlement over x402 or escrow. The merchant side learns none of them. A store owner installs a plugin, connects a payout destination, and their existing catalog, prices and orders become reachable by machine.

Inbound and outbound are separate problems and Facet solves both. Inbound, the Terminal verifies the calling agent's identity credential against the issuer's published keys before anything moves. Outbound, every response Facet returns is signed, so the agent and the merchant can both prove afterwards what was quoted, what was ordered, and what settled, without trusting Facet's word for it.

On 10 June 2026, in the first Agent-to-Business commerce transaction on x402B, an autonomous AI agent completed an end-to-end purchase from the real LARUCE Beauty Shopify store. It proved its identity, checked out, and the payment settled over x402B, Boson Protocol's escrow exchange mechanism, on Base mainnet. Funds were held in escrow until the order was fulfilled. Boson Protocol names Facet a launch partner for x402B.

It has been running on mainnet since, and the milestone that matters is not that an agent can pay. Plenty of demonstrations prove that. It is what happens when the deal goes wrong. Facet has validated the complete commercial lifecycle on Base mainnet, past the happy path: an agent buys, the order is cancelled, a dispute is raised, and the money is returned. Including the case almost nobody has closed, a contested order resolved to a partial refund that both the buyer and the seller cryptographically signed, settled on-chain.

That is the recourse layer, and it is the hard part of agent commerce. A payment rail with no path back is not commerce, it is a one-way transfer.

Read the writeup on the Devlog, inspect the integration in the open-source SDK, and request the narrated demo in the data room below.

McKinsey sizes agentic-commerce GMV at $3 trillion to $5 trillion globally by 2030, with roughly $1 trillion of that in the US. That is the market forming. The question for any company in it is distribution, and this is the part we did not have to solve from scratch.

Distribution is a plugin, not a sales team. Facet ships native to the two platforms that own independent commerce. WooCommerce is the primary channel, with roughly 6M live stores and the Facet plugin live today. Shopify is second, roughly 4M stores, with the app submitted and running the same rails. Every one of those stores is one install from live, with no outbound motion required to reach them.

Everyone is building the pay button. Facet is building the accept side. The pay button, giving an agent a way to spend, is crowded and commoditizing fast: platform checkouts, x402, AP2, browser agents. The accept side, letting an independent merchant actually take an agent's order and get paid for it, is where almost nobody is working, and it is the side that compounds.

An open protocol with a paid hosted layer. Five revenue lines run over the same rail, and a merchant can start on the free tier without a card. No token, and no governance dilution.

A fifth line, commerce analytics, sells per-seat dashboards and a data API over the agent traffic, conversion and settlement record a merchant generates by being connected at all.

Weston Le, Founder and CEO. Shipped Facet's acceptance layer into production solo and before raising: the Terminal, its own agent-identity issuer and verifier, signed response provenance, UCP checkout, the escrow settlement rail, and the WooCommerce plugin. Fifteen years operating in consumer packaged goods and food and beverage manufacturing, which is the merchant side of this market rather than the protocol side, and the reason the product is shaped the way it is.

Justin Banon, strategic advisor. Co-founder and CEO of Boson Protocol, the escrow rail Facet settles on. Previously scaled Priority Pass from $50M to $1B. Advises on settlement rails and go-to-market.

Greg Baker, strategic advisor. Partner at Rocicorp. Founding engineer at YES.com, acquired by Twitter. Gmail and Inbox tech lead at Google, and rebuilt twitter.com as a React progressive web app. Advises on scale and engineering access.

Hiring with the round: a founding protocol engineer, a founding frontend and SDK engineer, and founding go-to-market.

The adapter is live today. The seed funds the scale, not the science. Facet is raising an $8M seed and targeting a lead, having reached a live production money rail on no outside capital.

Dated milestones, the hiring plan, use of funds, and the operating model are in the data room rather than on this page. Request access below.