Is Your Store Actually Agent-Ready, or Just Agent-Ready in the Deck?
Short answer: probably not, and you can check in about five minutes without asking anyone's permission.
"Agent-ready" has become one of those phrases every commerce vendor puts on a landing page, the way "cloud-native" showed up on every pitch deck in 2015. It doesn't mean much on its own. What it should mean is specific and checkable: can an AI shopping agent read your catalog correctly, complete a transaction against it, and trust that what it read five minutes ago is still true. Most stores that claim the label pass none of the three.
1. Can an agent actually read your catalog?
Not "do you have a product feed." Everyone has a product feed. The question is whether that feed says anything an agent can act on.
We wrote up a version of this problem when we looked at semantic search on our own catalog: a vector search engine ranked sweatpants first for "breathable for a hot summer day," because every product description in the catalog was the same marketing paragraph with one noun swapped. A human skimming the page doesn't notice. An agent parsing attributes has nothing to parse: fabric, fit, and use case are all buried in the same sentence shape, repeated five hundred times.
The check: pull five product descriptions side by side. If they read like the same paragraph with the nouns swapped, no protocol integration fixes that. Structured data has to describe something structured to begin with.
2. Can an agent actually transact, not just browse?
This is the part most "agent-ready" claims skip entirely, because it's the part that requires real integration work, not a feed export.
Three protocols are shipping now: OpenAI and Stripe's Agentic Commerce Protocol (ACP), Google's Universal Commerce Protocol (UCP), and Retail MCP for agents that operate through Model Context Protocol servers directly. If your store runs on Shopify, you may already have an MCP endpoint without knowing it. A store that's genuinely agent-ready supports checkout through at least one of these, not just discovery. A store that's "agent-ready" in the marketing sense usually means its products show up in an AI Overview and nothing downstream of that actually works.
Ask a vendor which protocol they support, and ask them to show you a completed transaction through it, not a diagram of one. The diagram is easy. The transaction either clears or it doesn't.
3. Does the agent trust what it's looking at?
An agent that reads a price or an inventory count once and never checks again isn't shopping your store. It's shopping a memory of your store from whenever it last crawled. If your price or stock sync runs on a nightly batch job, an agent transacting at 2pm is working from data that's already wrong some fraction of the time. A human shopper might forgive a "sorry, that's actually out of stock" at checkout. An agent has no relationship with your brand to fall back on. It fails quietly and tries a competitor.
Real-time sync isn't a nice-to-have here. It's the difference between an agent completing a purchase and an agent silently deciding your store isn't reliable enough to recommend twice.
The five-minute check
Before you take anyone's word for it, ours included, there's a way to check your own store without booking a call. Run the free scan in your browser at mntfuture.com/open-source/agentready, or straight from the terminal:
npx @mntglobal/agentready your-store.comIt's free, and no email is required for the score. It runs 29 checks against exactly the things above: structured data quality, feed completeness, protocol support, and sync freshness. We built it and open-sourced it because we'd rather you had an objective read before any conversation with us or anyone else, not because we want to lead with a grade. If you run it and want to talk through what it found, that's a different, better conversation than a cold pitch based on a score we handed you unprompted.
What we actually ship here
This is exactly what our Agent-Ready Commerce work covers: an audit against the checks above, structured-feed work where the catalog itself is the problem, and integration with ACP, Google UCP, and Retail MCP where transaction support is missing. It's live work, not a roadmap slide. The protocols above are shipping now, not "coming soon."
Where we're building for ourselves, we say so. MnT Commerce, our own platform, is a demo catalog we dogfood these checks against. It has no live traffic or conversion numbers, and we don't pretend otherwise. LOBBI is the one platform in our portfolio running live in production with real users. If you check either against a case study claim, that's what you'll find, and we'd rather you found it from us than after.
Where to start
Run the scan first. It costs nothing and it's about your store, not a demo catalog. If it turns up gaps worth a real conversation, or if you'd rather skip straight to talking it through, that's what the strategy session is for: a senior consultant works through your data model, your protocol gaps, and a scalability plan, and you keep the written recommendation brief whether you hire us or not.