Skip to content
KalvyAI

Security and data handling

Careful with your systems, by default.

We work inside ERPs, stores and finance systems, and we build AI that can act in them. This page says exactly how we handle access, data and AI on every project.

01

Before we see anything

  • NDA first

    We sign your NDA, or send ours, before you share data, credentials or internal documents.

  • Only what the work needs

    We ask for access to the systems in scope, at the lowest level that does the job. Read-only first wherever that is enough.

  • A written scope

    The scope says which systems we touch and what we change, so there are no surprises in your production database.

02

Access to your systems

  • Named accounts

    Each engineer gets their own account, never a shared login, so every change can be traced to a person.

  • Your security tools

    We work through your VPN, SSO and two-factor sign-in where you have them, and follow your access policies.

  • Staging first

    We build and test on a staging copy. Production changes are planned, announced and come with a rollback plan.

  • Access ends with the project

    When the work ends, we hand over and you remove our accounts. We confirm in writing that we hold no copies of your data.

03

Your data

  • Stays in your systems

    Data is worked on in your environment or a staging copy you control, not copied to our laptops.

  • Anonymised when copied

    If a copy must leave your systems, for example for a test migration, personal data is masked first where the work allows it.

  • Never used to train AI

    Your data is not used to train AI models, by us or by the model providers we use for you.

  • Deleted when done

    Working copies, exports and test databases are deleted at the end of the project.

04

AI agents

  • Their own credentials

    Each agent has its own account with only the permissions its job needs. You can switch it off in one step.

  • Budgets and approvals

    Spend limits per agent, and a named person approves refunds, payments, price changes and deletions.

  • A full audit trail

    Every action is logged with the records it used, the decision and who approved it.

  • Providers you approve

    We send only what a task needs, to model providers you choose, with data retention turned off where the provider offers it. Self-hosted open models where data must stay in-house.

05

Code and ownership

  • Your repository

    Code lives in your repository and your cloud accounts from the first commit. You own it.

  • No secrets in code

    Passwords and API keys go in your secret store or environment settings, never in the repository.

  • Reviewed and tested

    Changes go through review and automated tests before they reach production.

  • Documented hand-over

    Runbooks, architecture notes and credentials lists, so your team or another partner can run it without us.

How we let AI touch your data

Five rules every agent we build follows.

AI that can act needs limits. These are built into the code, not written in a policy nobody reads.

  1. 01

    AI explains, code acts

    The model suggests and explains. Tested code makes the change, so every action is predictable.

  2. 02

    Always show the evidence

    Every AI answer links to the records it used. If it can't show its sources, it doesn't act.

  3. 03

    People approve risky actions

    Refunds, price changes, deletions and payments wait for a named person to approve them.

  4. 04

    Everything is logged or reversible

    Each action is written to an audit trail, with budgets and limits per agent. Most can be undone in one step.

  5. 05

    Say when unsure

    Low-confidence answers are flagged and handed to a person instead of guessed.

Our own systems

What runs on our side.

  • This website, our blog and our store run on our own server in Germany (EU), with HTTPS everywhere, a firewall, key-only server access and nightly backups.
  • Website analytics are self-hosted and cookieless; we do not use advertising trackers.
  • Form submissions are stored in our own database and used only to reply to you. See the privacy notice for details.

Found a security problem?

Email support@kalvyai.com with “Security” in the subject. We reply within two business days and fix confirmed issues as a priority. Please don't test against our clients' systems.

Need a security questionnaire or a data processing agreement filled in? Ask on the scoping call; we'll complete it before work starts.

Questions about security?

Ask on a 30-minute call with an engineer. We'll walk you through exactly how we would access your systems.

Book a scoping call Free AI-readiness assessment