Is your store invisible to AI shopping agents?
ChatGPT discovers products from feeds. Google's AI Mode transacts over UCP. Shopify ships MCP endpoints. Find out in seconds how your store looks to the agents: 29 checks, a letter grade, and exactly what to fix.
Prefer the terminal? npx @mntglobal/agentready your-store.com · Same engine, open source: about our OSS program
29 checks, in eight groups.
Every check is deterministic and every one shows its evidence, so you can verify the finding rather than take the grade on trust.
Structured data
Product, Offer and Organization markup, and whether price, availability and identifiers are actually present rather than declared empty. An agent that cannot read a price cannot quote you in an answer.
Agent access
What robots.txt allows, which AI crawlers are blocked outright, and whether the pages an agent needs are reachable without executing JavaScript it will not run.
Product feeds
Whether a machine-readable catalogue exists at all, how complete it is, and whether it is discoverable from where a crawler will look for it.
Protocol endpoints
Discovery for the agentic commerce protocols: ACP, Google's UCP and Retail MCP. This is the part almost every store fails today, and the part that decides whether an agent can transact rather than only describe.
Machine readability
Server-rendered content, clean canonical URLs, sitemaps that resolve, and llms.txt. The unglamorous plumbing that decides whether a crawl gets your catalogue or your loading spinner.
AEO citability
Whether your pages answer questions in a form an answer engine can lift and attribute: specifics, named entities, and claims that stand on their own out of context.
Data freshness
How recently the catalogue and its signals changed. Stale stock and stale prices are worse than missing ones, because an agent will confidently quote them.
Accessibility
The WCAG failures that also break machine parsing: missing names, unlabelled controls, content only reachable by mouse. Agents and screen readers fail on the same things.
What people ask about the grade.
It is the share of applicable points earned across 29 deterministic checks, mapped to a letter. It is not an opinion or a model's judgement: run it twice on an unchanged store and you get the same score. Anything the scan could not assess is excluded from the denominator rather than counted as a failure.
No. The scan makes roughly fifteen polite requests with a declared user agent, obeys robots.txt, and analyses what comes back statically. It never crawls your whole catalogue and never posts anything.
Not for the score. The grade, the category breakdown and every check with its evidence appear on screen. Email is only for the remediation plan, which is a written document a person puts together.
Because buying is starting to happen inside assistants rather than only on your site. ChatGPT discovers products from feeds, Google's AI Mode transacts over UCP, and Shopify ships MCP endpoints. A store that is invisible at that layer does not lose ranking, it loses the transaction entirely.
Yes, MIT-licensed. The same engine runs on the command line with npx @mntglobal/agentready your-store.com, so you can read every check, run it in CI, and point it at a staging environment before anything is public.
Almost always structured data and feeds, in that order: they are the cheapest to fix and the most widely consumed. Protocol endpoints matter more but cost more, so they are worth doing once the readable basics are in place.
agentready finds the gaps. We build the fixes.
Bring your report to a free strategy session: a senior consultant maps every finding to a concrete fix plan: feeds, schema, UCP/MCP integration, the lot.