Yours does not have to be. Paste your URL and see exactly which of the three layers you are missing, and how to fix them.
Runs server-side in a few seconds. Nothing is stored. Static sites can pass too. See a real setup →
Agents look for three things in order: a file that says you exist, a spec that says what you offer, and an endpoint they can call. Most sites have none of them, so agents guess, misquote, or move on.
A plain-text file at your root that introduces your site to an agent: who you are, what you sell, and where to look next.
↳ Agents find you instead of guessing from scraped HTML.
A machine-readable description of what your site can do, declared in llms.txt so agents read your catalog without crawling.
↳ Agents describe your offering accurately, not approximately.
A live server agents can actually call, with the right guards and a verified profile, so they can transact, not just read.
↳ Commerce agents can connect and act on your behalf.
As more people delegate research and buying to AI agents, agent-readiness is the difference between being found and being skipped. The three layers take an afternoon to add. Being invisible costs you every query you never see.
How a real storefront went from unreadable to fully agent-ready, layer by layer.
A Shopify coffee roaster added all three layers in an afternoon: an llms.txt, a valid OpenAPI spec, and a storefront agents can find and call.
Read the case study →Every check that fails comes with a plain-English fix. Here is how we close each gap.
We write an llms.txt that tells agents who you are, what you sell, and where they should look next. Real content in your brand voice, not boilerplate.
We publish a valid OpenAPI 3.1.0 file and declare it in llms.txt, so agents can read your catalog without crawling.
We stand up an MCP endpoint with the right guards and a verified profile, so agents can actually transact, not just read.
The check is always free and runs as many times as you like. If you would rather not touch it yourself, we will.
Paste a URL, get a score out of 100 across all three layers, plus exactly what is missing and how to fix it yourself.
We add every layer you are missing and verify your site scores green here. You approve, we ship.
A few seconds. We run the three probes server-side in parallel and return a score as soon as they answer.
No. The check runs server-side in a few seconds and nothing is saved. Run it as many times as you like.
Agents can describe you accurately, answer questions about what you offer, and (with an MCP endpoint) actually transact with you. As more people delegate research and buying to agents, that is the difference between being found and being skipped.
No. We check that the three layers are present and basically valid, not that every field is perfect. It is a fast readiness signal, not a full lint.
No. Probes identify as AgentReadyCheck/1.0 so you can allowlist them. When access protection answers instead of your file, we report it as "couldn't check", never as a failure.