Skip to content

Administering the platform's own environment

Administering the platform's own environment

On a multi-tenant deployment the platform root environment (cboxid.com on the hosted service, your root host when you self-host) is how every workspace owner and every operator signs in to Cbox ID. It is an environment like any other. It has sign-in methods, an authentication policy, social login, branding and a typeface, SMS, Radar, apps, APIs, keys, webhooks, log streams, roles, feature flags, fine-grained authorization, pipes, hooks and an audit log.

Its platform operators administer it, with the same console, the same actions and the same API a tenant's administrators use. That console is also where they administer the deployment itself: platform admin is its Platform area.

Where

A tenant's environment The platform's own environment
Console https://<environment>.cboxid.com/admin/… https://cboxid.com/admin/…, the same paths
Who the workspace's owners, admins and developers active platform operators
REST /api/v1/… on the root with Cbox-Environment: <slug> /api/v1/… on the root with Cbox-Environment: <root slug or id>
MCP the root's /mcp, environment argument the root's /mcp, environment argument
CLI cbox id use <environment> cbox id use <root slug>
Keys that environment's cbid_env_… keys, on its host the root environment's cbid_env_… keys, on the root host

Ways in, for an operator:

  • the account menu: Platform environment, or Platform admin for the Platform area;
  • the context switcher in the topbar: Platform › Platform, on every console;
  • ⌘K: type Platform and choose Open Platform;
  • Platform › Environments: Open console on the Platform root row.

Every page says what it acts on, above the page. On the platform environment that line is a warning: the platform's own environment … a change here applies to all of them.

Platform admin

The deployment's own pages are the first area of the same console, Platform, at cboxid.com/admin/platform:

Page Address What it is for
Workspaces /admin/platform/workspaces Every workspace on the install. Open one to walk its projects and environments; Suspend and Reactivate are on its row and its page.
Environments /admin/platform/environments Every environment and who owns it. Create one, point platform admin at it, and Provision admin to seed its first organization and owner.
Organization tree /admin/platform/organizations The organizations of the environment platform admin is pointed at, as the reseller → customer → sub-unit tree. Re-parent, suspend, and open one to start a support session (impersonation) as one of its members.
Operators /admin/platform/operators Who runs the install. Add or suspend an operator.
Usage /admin/platform/usage Sign-ins, users and tokens across every environment.
Queues /admin/platform/queues Whether the queue workers are running, with the job monitor at /admin/platform/queues/monitor.
Search /admin/platform/search One search across every environment. ⌘K searches the platform environment itself.

The line above each of these pages is the warning: Platform admin: the whole deployment … a suspension, a new environment or an operator change here applies across all of them.

It is the same shell as the rest of /admin: the rail, the context switcher, ⌘K, the breadcrumbs and the account menu do not change when you move between Platform and the environment's own areas.

Every write is the same action the operator API runs (/api/v1/platform/… on the root, and the root's MCP tools); the console address does not change the API's.

Who sees it. Active platform operators, on the root's own host. Anybody else gets 404 for every /admin/platform address, signed in or not, and a tenant's host has no such address at all. On a single-tenant install there is no environment console; the same pages, at the same addresses, are a Platform area of the console's rail for its operators.

Support sessions

On Platform › Organization tree, open an organization and choose Impersonate on a member (never an owner or admin), with a reason. You land on that member's console with a banner naming them and Exit impersonation. Platform admin and the platform environment's console answer 404 while it lasts. Exit impersonation, or the 30-minute limit, brings you back to Platform › Organization tree. A user of the platform environment itself can be impersonated the same way from Users & orgs › Users, and Exit impersonation returns you to the platform environment's console.

Old addresses

Platform admin used to be a separate console at /platform. Every old page answers 301 to its new address, with the query string kept:

Old New
/platform /admin/platform/environments
/platform/workspaces, /platform/workspaces/{id} /admin/platform/workspaces, /admin/platform/workspaces/{id}
/platform/customers, /platform/customers/{id} /admin/platform/workspaces, /admin/platform/workspaces/{id}
/platform/organizations, /platform/organizations/{id} /admin/platform/organizations, /admin/platform/organizations/{id}
/platform/operators /admin/platform/operators
/platform/usage /admin/platform/usage
/platform/queues /admin/platform/queues
/platform/queues/monitor, /platform/queues/monitor/job/{uuid} /admin/platform/queues/monitor, /admin/platform/queues/monitor/job/{uuid}
/platform/search, /platform/search/jump/{id} /admin/platform/search, /admin/platform/search/jump/{id}
/operator /admin/platform
/platform/security /account

Form posts to the old addresses are not redirected; a page rendered before the move posts to the new address the next time it loads.

A workspace owner who is not an operator gets 404 for cboxid.com/admin. They cannot name the root environment over REST or MCP either (not_found).

Turn on Google for cboxid.com

  1. Open the platform environment's console and go to Authentication › Social login.
  2. Choose Google. Who is it for? is already The whole environment.
  3. Copy the redirect URI the form shows into your Google OAuth client. It is on the root host: https://cboxid.com/sso/oidc/<id>/callback. That is the only URI to register.
  4. Paste the client ID and secret, then press Turn on Google for everyone.

The Google button is now on https://cboxid.com/login. Over the API, the same thing is signin.social.set with environment_wide: true, sent to the root with Cbox-Environment naming the root environment.

It also signs people in to every environment console

Environment administrators are platform-root accounts. Somebody who opens acme.cboxid.com/admin/users signed out goes through these steps:

  1. They are sent to the root's sign-in page, cboxid.com/login. It offers the platform environment's buttons and sign-in methods.
  2. They sign in. With Google, the provider calls back to the root's redirect URI. Radar, two-factor, step-up and the "last used" badge apply as for any sign-in.
  3. They are handed to acme.cboxid.com by the signed, single-use POST handoff. No token appears in a URL.
  4. They land on /admin/users, the page they asked for. That page is checked twice: it must be an environment-console page, and its signature must match the handoff. A foreign or tampered target is dropped, and the person lands on the console's home instead.

Signing out of acme.cboxid.com/admin signs the person out of cboxid.com too, and out of every other console that sign-in opened. A handoff whose sign-in has ended is refused, even if it was never used. The tenant's own users signed in on acme.cboxid.com are a separate context: the console's sign-out leaves them signed in, and theirs leaves the console open.

Somebody who signs in fine but administers nothing in that environment sees You don't administer this environment. The page says their sign-in worked and offers Switch user and their projects.

A tenant's own social providers are for the people who sign in to that tenant's apps, on the tenant's hosted pages. They are never offered for the tenant's /admin. Each Social login page says which of the two it is.

The guards

The platform environment is where the operators sign in, so a change that would leave an operator with no way in is refused. The refusal names who would be stranded and what to do first. This is enforced in the actions themselves, so the console, REST, MCP and the CLI all refuse the same changes:

  • requiring SSO for the whole environment, or for an organization an operator belongs to, while that operator has no active SSO connection to use;
  • disabling or deleting the SSO connection that operators who must use SSO sign in with;
  • turning off passkeys and magic links for an operator who has neither a password nor any other way in;
  • removing or turning off the social provider that is an operator's only way in.

Approvals. A token or key acting on the platform environment needs its person's approval for every write. Elsewhere, only critical actions need one. The actions and their declared danger do not change. Only the credential's approval threshold is raised on the root. The console uses its own step-up.

There is no surface that deletes the root's signing keys. cbox-id:keys:rotate always creates the new key before it retires the old ones.

When the owner is not an operator

The first-run screen and cbox-id:install make the installer the first operator. A deployment can still have an owner who is not one, for example somebody who signed up before the deployment was claimed. Grant it from a shell on the deployment:

php artisan cbox-id:operator:grant [email protected]
  • The person must already sign in to the root. The command never creates a person.
  • It links the operator record to that existing account. Their password, passkeys and second factor are unchanged.
  • It is idempotent: running it again changes nothing and says so.
  • It refuses to undo a suspension unless you pass --reactivate.
  • It is recorded in the root's audit trail as operator.granted.

On Kubernetes, run it in one of the web pods. The hosted service runs in the cbox-id namespace; replace <web-deployment> with the name of your web Deployment:

kubectl -n cbox-id exec -it deploy/<web-deployment> -- php artisan cbox-id:operator:grant [email protected]

The person can open https://<root>/admin, and platform admin at https://<root>/admin/platform, on their next page load.

The workspace console on the root host

cboxid.com without /admin is a workspace's console: projects, team, keys, billing, settings, the team's own sign-in to Cbox ID (Team sign-in) and its activity log.

Pages that set something an environment owns are not pages there any more. These include Social login, Sign-in methods, Branding, Domains, Apps, Webhooks and Roles. Each used to configure the workspace's own record in the root environment while looking like the product's settings.

  • An operator who opens one is sent to the same page of the platform environment's console, for example /social-sign-in → /admin/social-sign-in. Team sign-in is included.
  • Anybody else is told, at that URL, that it is set in each environment's console, with a link to the same page in each environment they may open.
  • A write to one of those pages answers 404.

Single-tenant installs are unchanged. There the root environment is the product, and its organization console is its administration — with platform admin as its Platform area, at the same /admin/platform addresses, for its operators.