Security and data · Written for your CTO or IT lead

Someone’s asked you to sign off on a vendor you didn’t pick.

Here’s everything you’d otherwise spend three weeks and a forty-row spreadsheet getting out of us: how access works, where your data lives, how code ships, who else touches it, and what we don’t claim.

Six areas, and what each one means in practice.

Access and identity

Per-person credentials at the minimum level needed, removed the day the engagement ends. No shared logins.

  • Every person gets their own credential to every system. We do not have a team account for anything, and we will not accept one if you offer it.
  • Access is granted at the lowest level that lets the work happen, and raised temporarily and deliberately when it cannot. A request to raise it is made in the shared channel, so you can see it.
  • Multi-factor authentication is mandatory on every account, ours and yours, including the ones opened for the project.
  • Credentials live in a password manager with per-project vaults. They are never in a repository, a ticket, a spreadsheet or a message.
  • Access is removed on the day an engagement finishes, not at the end of the month. Offboarding a person from Unflow removes every client credential the same day.
  • We keep a written record of who had access to what and when, and we will hand it to you on request.
Access and identity

Per-person credentials at the minimum level needed, removed the day the engagement ends. No shared logins.

  • Every person gets their own credential to every system. We do not have a team account for anything, and we will not accept one if you offer it.
  • Access is granted at the lowest level that lets the work happen, and raised temporarily and deliberately when it cannot. A request to raise it is made in the shared channel, so you can see it.
  • Multi-factor authentication is mandatory on every account, ours and yours, including the ones opened for the project.
  • Credentials live in a password manager with per-project vaults. They are never in a repository, a ticket, a spreadsheet or a message.
  • Access is removed on the day an engagement finishes, not at the end of the month. Offboarding a person from Unflow removes every client credential the same day.
  • We keep a written record of who had access to what and when, and we will hand it to you on request.
Your data

It lives in your own accounts, unless you’ve asked us to host it, and it’s never copied to a laptop. We work on anonymized or synthetic data.

  • The database, the file storage and the logs are in your cloud account, in a region you choose, from the first commit.
  • We develop against anonymized or synthetic data. Where a bug can only be reproduced on real data, we ask in writing, work inside your environment, and delete anything extracted when the ticket closes.
  • Production data is never copied to a laptop. Engineer machines are encrypted at rest with remote wipe, but that is a second line of defense, not the plan.
  • Personal data is minimized before it reaches anything it does not need to reach, including a model provider, where we redact before the call unless the task makes that impossible and you have agreed otherwise.
  • Where we process personal data on your behalf we sign a data processing agreement first, as processor to your controller, under GDPR and UK GDPR where they apply.
  • On handover, any copy of your data we hold outside your accounts is destroyed and we confirm that in writing.
Code, review and deployment

Pull-request review on every change, automated tests, CI, and a preview environment per change. Generated code is held to the same standard.

  • Every change is reviewed by a second engineer before it merges. There is no exception for founders and no exception for a hotfix; an urgent change gets a faster reviewer, not no reviewer.
  • Automated tests and continuous integration run on every change. A red build does not merge.
  • Each change gets its own preview environment, so what is tested is what ships.
  • Code generated with AI tooling is reviewed and tested identically to code written by hand. We remain responsible for it, and the warranty in our terms applies to it without distinction.
  • Dependencies are scanned continuously and updated every sprint, and monthly under a maintenance retainer. We follow OWASP practice for the classes of flaw that scanning does not catch.
  • Migrations are reversible and no destructive migration runs without a documented, tested backup step.
  • The repositories are in a GitHub organization in your name, from sprint one. Branch protection and required review are configured there, so the policy is enforced by your account rather than promised by us.
People

Everyone on the team works under written confidentiality obligations, on managed and encrypted machines.

  • Everyone who touches your systems is an employee or a long-standing contractor under a written agreement with confidentiality and IP assignment clauses.
  • Machines are encrypted at rest, screen-locked, and enrolled in device management with remote wipe.
  • You’ll know the names of everyone on your project, and we’ll tell you before that list changes.
  • Contractors from the vetted pool are onboarded to the same standard and offboarded the same day their sprint ends.
  • Security practice is part of onboarding and reviewed annually, not a certificate collected once.
Incidents

Acknowledged within one hour on a Priority retainer, 24/7. Every incident gets a written note, whether or not it was ours.

  • An incident is the system being unavailable or materially wrong for its users. We do not grade our own homework on whether something counts.
  • Priority retainer: acknowledged within one hour, 24/7, with best-effort resolution and hourly updates until it is closed.
  • Essential and Standard: handled in business hours, 8am to 6pm US Central.
  • A personal data breach affecting your data is reported to you within 48 hours of us becoming aware, with what we know at that point rather than a complete picture later.
  • Every incident produces a written note: what happened, what the impact was, what fixed it and what changes so it does not recur. You get it whether or not the cause was ours.
Continuity

Everything’s documented and, by default, runs in your own accounts, so the work keeps going with or without us.

  • By default, every account is opened in your name and billed to you, so there’s nothing to migrate at the end. If you’d rather host with us, we agree the rate in writing and hand the accounts over whenever you ask.
  • Documentation and a runbook are delivered at handover, whether or not you continue with a retainer.
  • No single person at Unflow is the only one who understands a client system: at least two people work on every project, and the review requirement is what keeps that true.
  • Backups are configured in your account with point-in-time restore, and we test a restore before launch rather than assuming one works.
  • If Unflow stopped existing tomorrow, your software would keep running and any competent TypeScript team could pick it up. That is a design goal, not a consolation.

Everyone who could touch your data, named.

This is what a typical engagement involves. Your proposal names the ones yours will use, and we tell you before that list changes.

Provider What for Region
Amazon Web Services Compute, storage, queues Your chosen region
Vercel Frontend hosting and previews Your chosen region
Neon Postgres Your chosen region
GitHub Source control and CI US
Anthropic / OpenAI Model inference, on your own account US or EU
Sentry Error monitoring EU
Google Workspace Email and documents EU

What we claim, including what we do not.

OWASP practice

Applied on every build, and the thing we would point to first. It is a practice, not a certificate, and we describe it as one.

In practice
GDPR and UK GDPR

We sign a data processing agreement as processor before any personal data reaches us, and a founder acts as data protection lead where an engagement needs one named.

In practice
SOC 2 and ISO 27001

Both are under way. We don’t have confirmed dates yet, so we won’t guess at them here. We’ll add them to this page once we actually hold the certificates.

In progress
HIPAA, PCI-DSS

We do not take on engagements that put us in scope for either. We will build alongside a compliant system, reading from it, writing to it, without handling the regulated data ourselves.

Not claimed

The questionnaire, answered in advance.

Where will our data live?

In your own cloud account by default, in a region you pick, from the first commit. If you’d rather we host it, we’ll agree that in writing and hand the accounts over whenever you ask.

Who at Unflow can access our production systems?

The named people on your project, each with their own credential at the lowest level that works, and only for as long as the engagement runs. We will give you the list on request, and you can revoke any of it yourself at any time because the accounts are yours.

Do you use offshore or subcontracted developers?

No. Engineering is our own team in Portugal. Where a sprint needs a specialist we use a vetted pool of people we have worked with, onboarded to the same standard and named to you before they start.

What happens to our data if we stop working with you?

Nothing has to move: it’s already in your accounts, or we hand those accounts to you. Our access is removed the day we finish, any copy we hold outside them is destroyed, and you get that in writing.

Do you use AI tools on our codebase?

Yes, for scaffolding, tests and first drafts. Every line is reviewed by a second engineer and passes automated tests before it merges, and we are contractually responsible for it exactly as we are for hand-written code. Your confidential material and personal data are not submitted to a model outside the arrangements in the data processing agreement.

Can we run a penetration test?

Yes, and we would encourage it. Tell us the window. We will fix anything material that is ours at no charge, and quote anything that is a change to scope rather than a defect.

What is your disaster recovery plan?

Backups with point-in-time restore in your account, and a restore tested before launch rather than assumed. The runbook you receive at handover describes how to do it without us.

Do you carry insurance?

Professional indemnity and public liability, appropriate to the size of engagement we take. A certificate is available on request before a project starts.

Got a question that isn’t on the list? Our CTO will answer it personally.