onrup

Docs

Security model

How does Onrup isolate and protect tenant data?

Every tenant’s data is isolated at the storage and database layer, keys carry least-privilege scopes, administrative actions require an interactive session, and consequential actions are written to an append-only audit log.

Isolation

Datasets, checkpoints, weights, logs and evaluation artefacts are stored under a per-tenant prefix, and access is scoped at the storage layer rather than only in application code. Identity and commerce records are isolated per tenant at the database layer as well.

Serving is tenant-scoped. Adapters can be co-resident with each other within your account, never across accounts.

Least privilege

Audit

Consequential actions — keys created and revoked, models deployed and retired, limits changed, data deleted — are written to an append-only log. Append-only is what makes it evidence: a log that can be edited cannot distinguish an absent record from a deleted one.

What we do not publish

We do not publish which compute providers we use, how capacity is selected, or how the platform is deployed. That is not evasion about security posture — it is the same reason your own architecture diagrams are not on your website.

What a caller can observe is publishable and is published: isolation guarantees, credential model, audit behaviour, rates and limits. Security questionnaires and review are supported on the Enterprise plan.

Frequently asked questions

Is my training data used to improve any model?

No. It is used for the runs you start and nothing else.

Do you have SOC 2?

We support security review and questionnaires on Enterprise. If a certification is a contractual requirement today, ask us directly rather than inferring an answer from this page.

Start with the free tier

A magic link creates your account, your tenant and your first API key. No card until you ask for compute.