AI you can actually trust with your data

The build itself, not a proof of concept.

• Your data stays yours

  Agents run against your systems inside your boundary. We do not pool your data with other clients and we never feed it into public model training.

• Access controls

  Role based permissions, least privilege service accounts and scoped API keys mean every agent only touches what it genuinely needs to do its job.

• Audit logs

  Every action an agent takes is recorded with who, what and when, so you can review, explain and prove exactly what happened at any point.

• Data residency

  Hosting in New Zealand where you need it, with encryption in transit and at rest, to keep you onside with local privacy expectations.

• Zero trust by default

  Nothing is trusted because of where it sits on the network. Every call is authenticated and authorised on its own, so a compromised component does not inherit the run of the system.

• Deterministic egress

  Outbound traffic goes to a known, declared set of destinations and nowhere else. An agent cannot quietly reach a service nobody approved, which is the failure mode that makes model driven systems hard to sign off.

• DLP filtering

  Personal information and export controlled material are detected and stopped before they leave the boundary, on the way into a model as well as on the way out.

• Per tenant keys

  Each client holds their own keys. There is no shared secret whose compromise reaches further than the one tenant it belongs to.

• Immutable audit logs

  The record of what an agent did is append only, so the log can be relied on in a review rather than being something an operator could quietly tidy up afterwards.