One read-mostly home to act as CEO: see the live system model, every role and the model it runs on, decide the changes awaiting you, and read spend & settings. Cost is live OPERATIONAL TELEMETRY (cumulative spend by role since the last reset) — visibility only, never a decision input. Autonomy values remain explicit placeholders pending their backend (CHG-B autonomy).
The single visual source of truth: organisation → column → role, rendered from the roles the engine already exposes (/roles, /models/roles). Click a column to zoom in; use the breadcrumb to zoom back out.
Connect to load the system model.
Every role as a card: name, column, the model it runs on, and "cost since reset" — the live cumulative spend attributed to that role/process step (operational telemetry, from /costs/by-role).
Connect to load roles.
Only the changes whose next legal move is your sovereign decision (approve / object / go-live), driven by legal_next. Each shows its problem and the latest role reasoning inline. The buttons call the SAME endpoints the portal uses.
Connect to load your decision inbox.
Cumulative API spend BY ROLE since the last reset (operational visibility only — never a spend-based decision input). "Reset" moves only the display baseline; the underlying spend facts are preserved on the immutable board for audit. Measurement window: spend recorded after the last reset timestamp.
Connect to load spend telemetry.
Read-only view. Autonomy (manual/auto) switches are governed by CHG-B's autonomy.yaml and shown as placeholders. Model assignments are read live from /models/roles. No silent writes.
| Manual / auto autonomy switches | — PENDING CHG-B |
| Budgets | — (not a spend-decision input; telemetry only) |
| Model assignments | view live → |