Dashboard|

Permissions

The scopes a Cencori OAuth app can ask for, and exactly what each allows.

A permission (a scope) names one kind of thing and a level: projects:read, projects:write. Write includes read, so projects:write is enough to list projects.

A token can only act on the one organization the person chose, and only on the things its permissions name.

ScopeAllowsWho can grant it
organization:readRead the organization's name, slug, plan and number of members.Owner or admin
projects:readList the organization's projects and read one.Owner or admin
projects:writeCreate a project and change a project's name or description. Includes projects:read.Owner or admin
api_keys:readList a project's API keys: name, prefix, type and when each was last used. Never the key's value.Owner or admin
api_keys:writeCreate and revoke API keys. Includes api_keys:read.Owner only
usage:readRead requests, tokens and cost for the organization or one project.Owner or admin

What api_keys:write means

A project API key can call models, and those calls are billed to the organization. An app with api_keys:write can therefore create something that spends the organization's money. This is why only an owner can grant it, and why the approval screen says so.

Ask for it only if your product needs to call models on the user's behalf, or to hand a key to something it deploys for them.

What no permission allows

OAuth tokens cannot:

  • call models or any other part of the AI API
  • delete a project
  • read the value of an existing API key
  • change members, roles, billing, plans or organization settings
  • see any organization other than the one chosen

Ask for less

Request the smallest set that does the job. A person is more likely to approve a short list, and a leaked token can do less.

Your app registers the most it can ever ask for. Each authorization request can ask for all of it or a subset:

scope=organization:read projects:read

A request for a permission your app did not register fails with invalid_scope.