Clear agent diagnosis — even for first-timers
A real agent walks signup → pre-checkout and shows where it stalls—with screenshots. Free, one-time diagnosis for consented sites.
Never completes a purchase · Never bypasses CAPTCHA · No run without consent
Zero knowledge? Start with the demo or an assisted request.
View the public demo, request assisted diagnosis, or run locally. Zero-knowledge users usually start with demo or assisted.
Only needed for local runs. Choose “③ Run the command yourself” to reveal the form; then enter confirmer, date, URL, customer ID, and authority. Demo and assisted paths skip this form.
Signup → browse → cart → pre-checkout. Purchase confirms are never clicked; CAPTCHA stops the run.
Agentrix Score, failure reasons, screenshots, and fix hints in one diagnostic document.
Authorization from the site owner (or rightful authority) is required. Zero knowledge? Pick ① demo or ② assisted. The command form only appears after you choose ③. The site run itself happens via assisted request or your local terminal.
You can start without programming. The easiest options are “view the public demo” or “request assisted diagnosis”.
The command form appears only after you choose path ③.
Internal/test URL runs are hidden from this list.
Live demo result
A real diagnosis against a public demo store. Reached signup, filled fields, then stopped on CAPTCHA without bypass — even Shopify’s Hydrogen demo stalls here.
How far can agents get on your site?
Agentrix Score
Being findable—and buyable—by agents is becoming a new kind of site reachability.
Stripe, OpenAI, Shopify and others are opening agent payment paths. Flows designed only for humans are no longer enough.
Based on public agent-commerce announcements and docs
Agents often abandon without an error page you would notice. Opportunity loss rarely shows up as a clear metric unless you measure it.
Typical stops in real runs: CAPTCHA, missing labels, forced signup
Not a security audit, not uptime monitoring—just whether an agent can get from signup to pre-checkout. Agentrix Score is that first yardstick.
Fixed 100-point score · consented sites only
Consent, safety limits, and score—one story that works for sales demos and legal review.
Without confirmer name and date, diagnosis cannot start. There is also no in-browser Run button.
Never clicks purchase confirm, caps at 40 actions / $2, stops immediately on CAPTCHA.
Fixed 100-point rubric — same yardstick every time, not gut feel.
The first diagnosis that helps a site become one agents can find—and buy from.
Show exactly where agents stall, with screenshots.
Share failure reasons like CAPTCHA stops and missing labels to prioritize fixes.
Consent records, a dedicated diagnosis bot name, and no purchase execution by design.
Agentrix measures whether an agent can get through. It is not a substitute for config reviews or vulnerability scans.
Swipe sideways to see the full table
| Agentrix | Security scanners & config checkers | |
|---|---|---|
| What it measures | Can agents reach signup → pre-checkout? | Config, code, known vulns, etc. |
| How it runs | A real agent walks a real browser | Scans and rule matching |
| Purchase | Never clicks purchase confirm | Out of scope (or separate process) |
| Deliverable | Score, why it stopped, screenshots | Finding lists and severity labels |
Yes. View the public demo, or request assisted diagnosis via Contact. Send consent and the target URL; we return a scored report. Local terminal runs are an optional developer path.
No. Without confirmer name and consent date, diagnosis cannot start. There is no web Run button. The consent form also writes an audit row when submitted.
No. Order/payment-confirm clicks and Enter/Space are refused; the run ends successfully at pre-checkout. Early Buy now CTAs are also never executed.
Runs stop immediately when CAPTCHA is detected. We never bypass. The stop is itself a measurement result.
Diagnosis reports are free for now. Anthropic API usage is billed to your key only for self-run terminal diagnoses. There is no monitoring or uptime guarantee.
Private by default. Only the public demo opens from the list. Individual reports use a share link—after you open it once, we keep access in a private browser cookie so the secret isn’t left in the URL. When sharing with teammates, use Copy link (it includes the token). Copying only the address bar often fails for the next person.
Reports are retained for 90 days from execution. Deletion requests via Contact are handled after identity checks, typically within 14 days.
No. It is a point-in-time agent-reachability check (signup → pre-checkout), not a penetration test.
Agentrix measures agent commerce readiness: whether an AI agent can walk signup through pre-checkout. It never completes a purchase and never bypasses CAPTCHA.
Agentrix does not shop for users. It diagnoses whether agents can traverse your flow and scores barriers (CAPTCHA, consent, purchase guards).
After the demo, request a diagnosis for your site. If you already have a report, continue the improvement conversation or re-run.
Send consent records and the target URL—we run assisted diagnosis and open an improvement conversation with an Agentrix Score report. We usually share a link within 2 business days.