Secrets is where you store the credentials your code needs, such as database passwords, API tokens and key files. You enter a value once, and PrivateMind delivers it by name to the workspaces and jobs that use it. So the value never has to sit in a notebook, a .env file, your version control or a chat message. This page explains how those stored values are protected.
How each secret is stored
Every secret is encrypted on its own. When you save a value, PrivateMind generates a new data key for that one secret and encrypts the value with it using AES-256-GCM, a standard authenticated-encryption algorithm. No two secrets share a data key.
The data key is then wrapped (encrypted again) by a key that belongs to your organisation. This is called envelope encryption: the wrapped data key can sit next to the encrypted value in the database, while the key that unlocks it is held somewhere stronger. Your organisation's key lives in a dedicated key management service that we run as part of the platform, and it never leaves that service.
For each secret, the database holds the encrypted value and the wrapped data key. The value and the key that unlocks it are never stored together in a usable form. A copy of the database, or one of its backups, can't be decrypted without your organisation's key, and that key isn't in the database.
Bound to its own identity
Each stored secret is cryptographically tied to its identity: the organisation it belongs to, the secret's own identifier, and its name and owner. The binding is built into the encryption, so it can't be peeled off and reattached.
As a result, a secret decrypts only in the exact place it was created. If a record is copied into another organisation, swapped onto a different secret or renamed, decryption fails and the value can't be read.
Who can use your organisation's key
Only the PrivateMind application can use your organisation's key, and only to unwrap the data key of a secret that belongs to your organisation. To read a secret, the application gets a short-lived credential, scoped to your organisation's key alone and revoked automatically as soon as the operation finishes. One organisation's credential can't reach another organisation's key.
The key can't be exported, copied or backed up out of the key management service through the application. No application path can take a plaintext copy of it away.
Keys rotate
Your organisation's key is rolled to a new version on a regular schedule, and it can also be rotated on demand. From the moment a new version exists, every new secret is wrapped under it.
Secrets you stored earlier move to the new version automatically, in the background. During this re-wrapping the value is never decrypted to plaintext and the stored encrypted value doesn't change: only the wrapping around the data key is refreshed. Your secrets stay available throughout, and you don't need to do anything.
Once existing secrets have moved, an old key version can be retired, and then it can no longer decrypt anything.
Values are never shown again
After you enter a secret, PrivateMind never shows it again, to anyone. You can see the names of your secrets, who owns them, any note you added, where each one is attached and when it was last delivered, but never the value. To change a secret, replace it with a new value. You can't read the old one back out of the app.
Changes to secrets, and each delivery of a secret to a workspace or job, are recorded in your organisation's audit log by name, never by value.
Delivery to workspaces and jobs
When you attach a secret to a workspace or a job, PrivateMind delivers it as a read-only file that your code reads by name. The delivery is encrypted in transit, so the credential appears in plaintext only inside the compute environment that runs your workspace or job. It never appears in plaintext in the systems that route it there.
What Secrets protects, and what it does not
Secrets is built to keep credentials out of the places they leak by accident: notebooks, source files, version control, chat and the application's own screens.
It doesn't hide a value from code running in the workspace where you attached the secret. Anything that can run code in that workspace can read the delivered file: anyone who can open the workspace, every program they run and any coding agent working there. This is deliberate: you attach a secret so your code can use it, and code that can use a credential can read it.
Treat a workspace as trusted by everyone who can open it. Attach a secret where your own code needs it, and share a workspace only with people who are allowed to use the credentials attached to it. Secrets can't let some people in a shared workspace use a credential while hiding it from others.