Control who can read and change environments

Readers and writers per environment, protected production, change approvals, access requests, inheritance and the audit log.

Web
On this page
  1. Roles
  2. Protect production
  3. Ask for access
  4. Inherit from another environment
  5. Hints and audit

In Secrets, each project has environments (for example development, staging, production) and each environment has its own key. Only the people granted that environment can decrypt it.

Roles

  • Owner / admin of the project: manage members, environments and access.
  • Writer on an environment: read and change its variables.
  • Reader: read only. No access: the person sees that the environment exists, not its values.

Set them in the project's Access tab. Removing someone offers to rotate the keys, so they can't read anything added later; their machine tokens stop working at once.

Protect production

In the project's Settings, choose an environment's Protection:

  • Changes need approval — writers click Submit for approval; another project admin opens Change requests → Review and approves or rejects. Nobody approves their own change.
  • Admins only — read-only for everyone except project owners and admins.

Only project owners can lower a protection.

Ask for access

A member without access opens the environment and clicks Request access (read or write, with an optional reason — never paste a secret there). Admins see the request in the Access tab and Grant or Deny it.

Inherit from another environment

Settings → Inherits from makes an environment start from another one's variables (for example staging from shared). Inherited variables show as from shared; Override to set your own value. Exports, the CLI and the SDKs merge them the same way.

Hints and audit

Aman points out likely problems next to the variables: placeholder or very short secrets, production keys in an unprotected environment, machine tokens stored as values. The project's Activity tab lists every read — by people and machine tokens — with time, address and client; a token used from a new address is flagged and its creator gets an e-mail.