Security
Orme Innovation Inc. · Effective August 3, 2026
How Orme OS protects customer data, stated concretely. Everything on this page describes mechanisms that exist and are tested in our continuous integration, not intentions.
Isolation
Every customer workspace is isolated three ways at once: database row-level security on every tenant table, explicit tenant checks in the application layer, and composite foreign keys that make a cross-tenant reference unrepresentable. Fifteen automated isolation checks run on every code change and block the merge on any failure.
Human control
No message reaches anyone outside a customer's company without a named person approving the exact rendered content. Approvals are cryptographically bound to what was approved: editing a draft after approval invalidates the approval. Human corrections to the record are final; the system never overwrites them.
Encryption and credentials
Traffic is TLS in transit and encrypted at rest by our database provider. OAuth tokens and webhook secrets are additionally sealed with AES-256-GCM before storage. Provider credentials never appear in application logs, which carry identifiers rather than content.
Audit
Every write to a customer's record, by an agent or a person, produces an audit entry with the actor, the change, and the evidence behind it. The audit log accepts no updates or deletes from the running system.
Honest gaps
We do not yet hold a SOC 2 attestation; a Type I engagement is planned once our first design partners are live. If you find a vulnerability, email us and a person will respond quickly. We prefer an uncomfortable report to an unreported hole.
Contact: rajath@orme.co · 18 South 2nd Street, San Jose, CA 95113