Glossary
Short definitions for Agentrix, agent commerce readiness, and related terms—written for search and answer engines.
Quotable summary
- Findable is not buyable. Agentrix measures, with the store's consent, whether an AI agent can walk an EC site from signup to pre-checkout.
- What it measures: signup, login, product browse, add-to-cart, and reaching the pre-checkout screen — a real browser walk, not a static scan.
- What it does not measure: purchase completion, CAPTCHA bypass, vulnerability scanning, or how visible a store is in search/AI answers.
- A CAPTCHA stop is not a failure. It is the finding itself: this site's bot defense stops agents.
- Connecting an admin panel to AI (MCP-style admin integration) is not the same as agents being able to walk the storefront. One automates operations; the other is purchase-path reachability.
- Agentrix
- A diagnosis product that measures whether an AI agent can walk a site from signup through pre-checkout. Never completes a purchase; never bypasses CAPTCHA.
- Agent commerce readiness
- How completely an agent can finish critical flows (signup, login, browse, cart, pre-checkout). Quantified by the Agentrix Score.
- Agentrix Score
- A 0–100 reachability score from step success/failure and barrier tags. Mid-step success is credited only with observable evidence.
- Pre-checkout
- Reaching order/payment review before confirm. Does not include placing or authorizing payment.
- Consent check (consent gate)
- Diagnosis cannot start without the site owner’s (or rightful authority’s) name and consent date. This blocks unauthorized runs.
- No-purchase rule (purchase guard)
- The diagnosis bot never clicks order confirm, payment confirm, or Buy now. Reaching the review screen ends the run successfully.
- CAPTCHA stop
- Immediate halt on bot-challenge detection with no bypass. The stop itself is a reachability result.
- Customer ID
- A short alphanumeric id that distinguishes who requested the run (e.g. my-shop). Re-running the same customer ID and URL unlocks score history. Not used on the public demo.
- Score badge
- A small SVG image of the Agentrix Score for sites or decks. Not yet available: there is no way for a customer to opt in yet, so badges 404 on ordinary reports. When it opens, only opt-in reports will be served publicly; private diagnoses return 404. A separate time-limited Walk proof (`/api/badge/{id}?kind=walk`) is issued only when cart was observed, Walk is at least 80, and the scan is under 90 days old. It is not a permanent Ready mark.
- Filmstrip
- A horizontal strip of screenshots from the run. Public demos may omit captures.
- Reach / Read / Protocol
- The official frame. Reach is UI walkability measured today (Walk = commerce before bot defense; Policy = an intentional CAPTCHA or bot stop). Read is statutory pages and structured data; Protocol is UCP and similar. Read and Protocol are unchecked in this rubric — not scored as zero. Old reports are never rescored.
- Barrier ID
- A public vocabulary for stop reasons (e.g. JP-FORM-FURIGANA, JP-POLICY-CAPTCHA). Detected tags map onto IDs; the catalog lives at /en/barriers.
- Walk proof
- A time-limited attestation that an opted-in store was observed through cart, scored Walk at or above the threshold, and was scanned within 90 days. Not a guarantee that a purchase can complete.
- Agentic Commerce
- The idea of AI agents assisting or completing shopping flows. Agentrix diagnoses whether those flows are agent-passable; it does not shop for users.