AI & MCP

MCP security

How tokens, scopes and workspaces keep an AI tool to exactly what you allowed.

Tokens

  • A token is a long random secret starting with alara_pat_. It is shown once, when you create it.
  • Alara Space stores only a one-way hash of it, so it cannot be recovered or shown again, even by an administrator.
  • Tokens expire after 30, 90 or 365 days, or never, as chosen at creation. Revoking one takes effect on the next request.
  • Any problem with a token (missing, wrong, expired or revoked) gets the same answer, so the response gives nothing away.

What a token can reach

  • One workspace. A token is tied to the workspace it was created in. Spaces, pages, comments and files in any other workspace do not exist as far as it is concerned.
  • One person. It acts as the user who created it, and can do what that user can do in the workspace, nothing more.
  • Read or write. A read-only token is not offered the tools that change content, and is refused if it tries one anyway.
  • No destructive shortcuts. There is no tool to delete permanently: trashed pages can be restored, and replacing a page saves its old content as a version first.

Accountability

Everything a token writes is recorded as the token's user: page history, comments and the workspace audit log all show who made the change, exactly as if it had been made in the app.

Good practice

  • Create one token per tool or device, named so you can recognise it later.
  • Use read only unless the assistant needs to write.
  • Prefer an expiry; 90 days is the default.
  • Keep tokens out of shared files and source control. If one leaks, revoke it and create another.
If you leave a workspace
Revoke your tokens for it. A token keeps working until it expires or is revoked, so administrators offboarding someone should ask them to revoke theirs, or pick short expiries.