Product & engineering
Ship every week without breaking what makes money.
In the pilot, agents try your key journeys on each release, as a shopper on a phone, a keyboard and screen-reader user, and an AI shopping agent. We're building the hand-off that sends blockers to your coding agent and runs the check again.
Sign-up opens shortly.
The loop
Release, check, fix, re-check, confirm.
- 01
You release
A deploy, or a Shopify theme publish.
- 02
Agents try key journeys
Three kinds of customer, stopping before payment.
- 03
Blockers found, or none found on this run
With a screenshot and the change to make. Never “works”.
- 04
Your agent fixes it
Through MCP to Claude Code or your own agent.
- 05
Re-check, then real visitors confirm
The journey runs again; live traffic settles it.
- Release check: key journeys tried on every release, stopping before payment Pilot
- Your coding agent gets the fixes Building
- Set it up from Claude Code Building
- The check runs itself after a fix Building
- Every change checked against your other pages: worked, didn't, or can't tell yet Building
What we promise, and what we don't
A blocker found is a fact. “None found on this run” is all we say otherwise.
A test run can show that something blocked a customer. It can't prove a site has no problems, so we never say a site or a checkout “works”, or that it's “accessible”.