Machine-readable commerce

A shopper asks an agent to buy something. The agent compares what it can read — price, stock, shipping, return terms, a checkout it can execute — and buys from whoever exposes them. A shop that only renders HTML for human eyes doesn't lose on price. It never enters the comparison. This is the 2005 SEO moment with the meter already running, and the merchant side of that rail takes a cut of everything that flows through it.

The agent, shopping watch which shop it can't see

What this is worth, per shop

Agent traffic
×3 / yr
and compounding
Shops ready
<4%
of the long tail
Take rate
0.4%
on agent orders
Why a toll and not a subscription: a shop will argue about €29/mo forever and will not argue about a slice of a sale that would not have existed. The pricing aligns us with the only number the merchant cares about, and it scales with the category instead of with our sales effort.

Agent-readiness of brassandoak.store

of 100 ·
Every red line is a reason an agent silently skipped this shop. None of them are visible in the shop's analytics, because an agent that can't read you never arrives to be counted — which is precisely why the free scan sells the product.

The generated feed one file, served at /.well-known/


        
Deliberately boring and open. The format has to look like something an engine would adopt, not like our proprietary lock-in — the money isn't in owning the schema, it's in being the thing that generates, hosts and keeps it true for a million shops that will never write JSON by hand.

Honest position

Who's moving: OpenAI shipped Instant Checkout with an open Agentic Commerce Protocol (with Stripe), Google has AP2, Shopify and Etsy wired themselves in early, and Cloudflare is building the pay-per-crawl layer. So the rail is real and being laid by giants — this is not a bet on whether it happens.

The gap we'd take: every one of those serves the head of the market. Shopify merchants get it free; the millions of shops on WooCommerce, PrestaShop, custom PHP and regional platforms get nothing, and they're the ones who can't write the integration themselves. A fleet that can crawl a shop, infer its catalogue and emit a correct feed without the merchant doing anything is the long-tail onramp.

The risks, plainly: the protocols are young and may consolidate into one we don't control, which turns us into a commodity generator; the platforms could absorb the long tail with a free plugin at any moment; and a take-rate model means touching money, which drags in exactly the KYC and merchant-of-record problems we otherwise avoid — the honest first version is a flat fee, not a toll, until that's solved.