Security & PCI
Card data stays on the terminal. Commands need keys we do not hold.
TickFlow Field is designed to stay out of your cardholder data environment and to make a remote fix as accountable as one done in person. Here is what that means for you; the mechanisms, for your security team and your QSA, are in our technical briefing.
PCI DSS scoping
Designed to keep you out of PCI scope.
Field never receives card data: it is removed on the device, and the server quarantines any bundle where a rule still matches. Field never touches the payment kernel, the card reader’s secure path or PIN entry. On Android, the agent is a regular or vendor-approved app and must not weaken the terminal’s PTS posture.
We write this scoping argument down for each customer’s QSA. The assessment and its conclusion stay with your QSA; Field holds no PCI certification of its own as a product.
Redaction on the device
Card numbers are gone before anything is sent.
Card numbers, track data, PINs and personal data are removed on the terminal, by the same rules on every kind of device. The server checks every report again and holds back anything that still looks like card data. Order numbers and other look-alikes survive, so the evidence stays useful.
Approvals
Risky fixes need two people. The service alone cannot run them.
Everyday fixes can run under your policy. A reboot or a wipe also needs an approval that the part of Field sending commands cannot produce on its own, so even a compromised server cannot make a terminal do something drastic.
Rolling out Hardware-backed device attestation: where the terminal supports it, a device proves it is genuine before it is trusted.
app.clear_dataon Suburbs T-11Requested by Uma U. — who cannot approve her own request.
Nobody approves their own request.
The person who asks for a destructive action cannot approve it. By default one approver is from your organisation and one from Orthoplex, and an approval expires and is checked again if the device changed in the meantime.
Remote sessions Rolling out
A last resort, with the door locked behind it.
Most problems are fixed by a named playbook. For the few that are not, an engineer can see the screen, take control or open a locked-down console. Each is a separate permission, and every one of them follows the same six rules.
- ViewSee the screen with the person at the device.
- ControlTap and type for them, when a playbook cannot.
- ConsoleA locked-down command console, for the cases nothing else covers.
- 01Off by defaultNobody can start a session until your organisation turns sessions on, per fleet.
- 02Consent on the deviceThe person at the terminal sees who is asking and why, and can say no.
- 03Approved firstEvery session is approved before it starts. The command console needs two people.
- 04Blank during paymentsThe screen goes dark to the viewer while a payment is in progress. Card entry is never seen.
- 05RecordedWho watched, what they did and for how long, kept with the ticket and in the ledger.
- 06Time-limitedA session ends on its own. Extending it means asking again.
The ledger
Every action leaves evidence you can verify.
Proposal, policy decision, approvals, the command, the device’s acknowledgement, the result and the verification: each is written to a tamper-evident ledger. Nobody can quietly change or remove an entry, you can check that the record is intact, and you can export it for your auditors.

- #18,229proposalapp.restart v3 · T-06 · triage 78%e57ae585…e70b
- #18,230policyruns on its own · change · known issueprev e57ae585…e70b → 5b399308…8741
- #18,231commandissued · signedprev 5b399308…8741 → 0f73dbee…79b0
- #18,232ackT-06 accepted · waited 4 s for idleprev 0f73dbee…79b0 → 594942d3…7507
- #18,233resultok · exit 0 · 3.1 sprev 594942d3…7507 → c87c69c8…06af
- #18,234verifyhealth check passedprev c87c69c8…06af → ecc800cf…146b
chain verified · 18,234 rows · no entry can be changed or removed
Our rules
Six rules decide every trade-off.
Dormant by default
No streaming. The device keeps its own recent evidence and sends only a small signed check-in until someone reports.
Redact at the edge
Card numbers, track data, PINs and personal data are stripped on the device. The server re-checks and never relies on itself alone.
Outbound only
Terminals sit behind merchant firewalls. The device always opens the connection; nothing ever connects to it.
Nothing arbitrary
Remote actions are named, signed playbooks with typed parameters. No shell. A remote session is a consented, recorded last resort.
Four-eyes
AI proposes, policy decides, people approve. Nobody approves their own request; destructive actions need two people.
Tamper-evident ledger
Every command, approval and result is a row in an append-only, hash-chained ledger you can verify and export.
Privacy & residency
Stored in Canada. Kept only as long as it is useful.
- Diagnostic bundles
- 30 days
- encrypted, then deleted
- Redacted search index
- 13 months
- for clusters and known issues
- Action ledger
- 7 years
- audit evidence
- AI providers
- No training
- zero or limited retention; you can switch AI off
Canadian storage by default, with a per-tenant region on Enterprise. Data minimisation is built in: dormant agents, redaction at the edge, no request bodies in network logs.
Bring your security team to the demo.
We will walk your security team and your QSA through the scoping argument, the redaction and the ledger, on a live terminal.
Or write to hello@tickflow.co