tenant_id vs sub: where to put tenant context in a JWT

Why conflating tenant identity with user identity in the sub claim is a multi-tenant anti-pattern, and the conventions that work instead.

Private: Decoding, analysis, signature verification, and token generation all run in your browser. Nothing is sent to any server, except a JWKS URL you explicitly enter and click Verify on.

At a glance

Anti-pattern
sub: "tenant_X:user_Y"
Recommended claim for active tenant
tenant_id, org_id, or workspace_id
Recommended claim for memberships
tenants[] or memberships[]
sub purpose (per RFC 7519)
Stable identifier for the user across sessions

A surprisingly common pattern in multi-tenant JWTs is to encode the tenant in the `sub` claim alongside the user id (e.g., `sub: "tenant_acme:user_42"`). It feels efficient — one claim, two pieces of data — but it conflates two unrelated concepts and breaks as soon as a user belongs to more than one tenant. Use a dedicated tenant claim instead.

Why the anti-pattern is tempting

The "compress two ids into one claim" approach has surface appeal. Every claim takes a few bytes, and JWT size matters when tokens go in headers. Conflating sub with the tenant feels like a way to save space and avoid adding a claim.

In practice, the savings are negligible (tokens are typically already 500–2000 bytes; one extra claim adds ~30) and the costs are real. Tooling that consumes JWTs — Sentry user binding, audit logs, analytics — all expect `sub` to be the user. Splitting it across tenants makes everything downstream more complex.

The conventions that actually work

Keep `sub` as the user id, period. Use `tenant_id` (or the provider-specific equivalent — `org_id`, `workspace_id`, `team_id`) for the active tenant. If users can belong to multiple tenants, list memberships in a `tenants[]` array of `{ id, role, permissions }`.

For permissions and roles, use `tenant_role` (or `org_role`) and `tenant_permissions` (or `org_permissions` / `permissions`). All of these claims should be scoped to the active tenant — the JWT represents one session in one tenant, even if the user has access to others.

Frequently asked questions

What's wrong with putting the tenant in sub?
Two problems. First, `sub` is supposed to be a stable user identifier — your tooling, audit logs, and analytics will treat it as such. Encoding the tenant means the same user looks like different people across tenants, breaking deduplication. Second, when a user joins a second tenant, you can no longer represent them with one `sub` — you have to issue different tokens with different `sub` values. Authentication becomes session-scoped, which is fine, but anything that ties to user history (rate limits per user, recent activity, etc.) breaks.
What if my JWT was already designed this way?
You can transition by adding a parallel `tenant_id` claim and switching consumers to read from it. The tenant-encoded `sub` becomes a vestigial detail you can phase out once nothing reads it. Keep `sub` as the literal user id going forward.
Should the tenant claim be the tenant id or the slug?
Use the id (a stable identifier) as the primary claim. Add the slug as a separate claim if you want it for display — slugs can change when a customer renames their workspace, and you don't want every active token to invalidate when that happens.

Related guides

Also useful: UUID GeneratorJSON Formatter