Setting Up Multi-Tenant Teams

One Coder server can run many independent OpenFlows teams. Each team is a tenant, isolated from the others by both identity and coordination state. This guide walks through adding additional teams and managing the set.

How one server serves many

• Each tenant is a distinct Coder user, a repository binding, and its own orchestrator workspace.
• Tenants share the same Coder server and coordination store, but never see each other's work.
• Isolation is enforced two ways at once: separate identity and access control, and separately namespaced coordination state. Even a misconfigured access control can't read another tenant's data.

Steps to add another team

1. Have the control plane running (from the Quickstart).
2. Add the tenant for the new team's repository. This creates the tenant's user and its orchestrator workspace, and initializes a fresh agent team.
3. Complete GitHub OAuth for that tenant, scoped to its repository.
4. Optionally configure notifications for that tenant's escalations.

That's it — the team is live and can start taking issues for its repository.

Managing the team set

From the control panel (or the equivalent command surface):

• List the tenants running on the server.
• Add more tenants as above.
• Clean a stuck tenant — reset stale or failed work to the start so its team can proceed.
• Remove a tenant and tear down its fleet.

What you should see

• Each tenant's status is visible separately; you can tell whose tickets are which.
• Teams operate in parallel without cross-talk.
• A cleanup on one tenant never disturbs another.

Isolation guarantees, restated

• Identity plane: each tenant is a separate user; access is controlled by role-based access control.
• Data plane: each tenant's coordination records live in its own namespace, structurally unreachable by other tenants' components.
• These compose into strong, defense-in-depth isolation on shared infrastructure.

Practical guidance

• Give each team a distinct, well-scoped repository binding so work doesn't collide downstream.
• Set notifications per tenant so escalations reach the right people.
• Use list regularly to keep an eye on the whole fleet across teams.
• Reserve clean and remove for deliberate operations; they reset or tear down a team's work.