Help

Help

How to read a report, what each finding means, and how to get our agents through your bot protection.

• Getting a human

  Support is [email protected] and reaches a person. Including the store URL, the run in question and what you expected instead of what happened turns a first reply into an answer rather than a request for more detail, because most questions about a report can only be answered against the specific run.

• What we store, and for how long

  A scan keeps what is needed to justify its findings: the pages as they were fetched, a trace of what the agent did in what order, and the cost of the run. Evidence is kept because a finding you cannot check is a claim rather than a result, and everything stored is drawn from a public storefront rather than from anything private.

• Billing, invoices and cancelling

  Where you subscribed decides which rail bills you: a subscription started inside the Shopify app is billed by Shopify, and one started on the website is billed by card. A storefront carries one active subscription at a time, which is enforced in the database rather than by the interface, and cancelling stops future scans without deleting the reports you already have.

• Plans, cadence, and what actually changes

  Plans differ mainly in cadence: how often a full agent journey runs against your store and how quickly a regression is caught. Some tiers also unlock additional surfaces, but the ladder never takes a surface away as the price rises, and a tier that alerts you to a finding always lets you read it.

• Connecting a Shopify store

  Installing the Shopify app lets a scan be enriched with data the storefront does not expose publicly, and it is never required. The scanner cannot know whether the app is installed, which is deliberate: a scan measures what an agent arriving from outside can see, and privileged data would make that measurement dishonest.

• Scan limits, and why they exist

  How often a store can be scanned is capped, and the cap is counted against the store being visited rather than against the account requesting it. That is deliberate: the limit exists to protect a merchant from us, so a cap that could be reset by asking from a different account would not be a limit at all.

• Allowlisting our agent

  Our agent declares itself in a published user agent, reaches your store from a fixed set of published egress addresses, and respects your robots rules. That means it can be admitted by verified identity rather than by turning bot protection down, which is a much narrower opening and the one your security team is likely to accept.

• Why a finding was not surfaced

  A report withholds a finding when the evidence does not support it. The three usual reasons are that a serious finding needs two concurring signals and only one was seen, that a change between scans is smaller than the measurement noise, or that the attempt stopped before the stage in question. Each is a deliberate choice to under-report rather than to guess.

• Reading your report

  A report reads as the story of one purchase attempt: four stages in the order the agent met them, each either passed or stalled, with the evidence for the verdict attached. Findings are ordered by where they occurred rather than by severity, because a failure early in the journey makes everything after it untested rather than fine.

• What a scan does, and what it does not do

  A scan sends a real AI agent at your storefront and records where it got to: whether it could find your catalogue, read your product data, compare you against an alternative, and drive your checkout toward the payment step. It stops before payment, it never places an order, and it reports what it observed rather than what it guessed.