Bounded Autonomy: Credit Budgets, Quiet Hours, Allowed Channels, and Kill Switches
How to configure workspace governance, role-based access, and three graduations of AI autonomy (`draft_only`, `schedule_with_confirmation`, `autonomous_schedule`).
- Autonomy is not a binary on/off switch; teams need graduated modes (`draft_only`, `schedule_with_confirmation`, `autonomous_schedule`) as trust matures.
- Hard operational guardrails—allowed channels, maximum cadence, quiet hours (`22:00–07:00`), and monthly credit budgets—prevent runaway automation.
- Multi-brand agencies require strict workspace isolation and role-based permissions (`Owner`, `Admin`, `Member`, `Viewer`) plus an instant stop switch.
Why Unbounded "Full Autopilot" Breaks Trust
Many AI tools promise "100% hands-off autopilot" with no middle ground. In reality, a regulated brand or client agency cannot let an agent post at 3:00 AM, burn through unlimited generation credits over a weekend, or publish unreviewed visuals. Without explicit boundaries, teams end up disabling automation entirely.
All-or-nothing autopilot + shared credentials across client brands + no spending cap or quiet hours.
3 Graduated Autonomy Modes + Quiet Hours & Credit Caps + Per-Workspace RBAC + Instant Strategy Stop Switch.
1. The Three Graduated Autonomy Modes
A mature brand operating system lets you dial autonomy up or down per workspace (`AG05`):
Mode 01 — `draft_only`: Agents and generators can research, analyze cohorts, and save drafts in the workspace, but cannot queue or schedule anything. Mode 02 — `schedule_with_confirmation`: Agents prepare complete drafts and propose optimal calendar slots, pausing at the confirmation gate for human sign-off. Mode 03 — `autonomous_schedule`: Within explicitly approved strategy rules, allowed channels, and cadence limits, delegated scheduling proceeds automatically while still respecting quiet hours, media review gates, and live-publish blocks.
GOVERNANCE RULEEven in `autonomous_schedule` mode, hard guardrails—quiet hours, allowed target channels, credit budgets, and media review—always take precedence.
{
"workspace": "fieldwork",
"autonomyState": {
"mode": "schedule_with_confirmation",
"allowedChannels": ["instagram", "facebook", "x", "linkedin"],
"maxPostsPerWeek": 6,
"quietHours": {
"enabled": true,
"start": "22:00",
"end": "07:00",
"timezone": "Europe/Warsaw"
},
"monthlyCredits": 1000,
"humanMediaReviewRequired": true,
"stopped": false
},
"workspaceRoles": {
"owner": "Full strategy, billing, autonomy & kill-switch control",
"admin": "Approve strategies, confirm publications & manage channels",
"member": "Create and edit drafts, research & reply suggestions",
"viewer": "Read-only access to calendar, drafts & cohort analytics"
}
}2. Hard Guardrails: Quiet Hours, Credit Budgets, and the Kill Switch
Four deterministic guardrails keep delegated work predictable: (1) Quiet Hours (`AG06`) block automated scheduling or replies during overnight windows (e.g., `22:00–07:00`); (2) Monthly Credit Budgets (`AG07`) cap tracked AI generation so a loop never exceeds allocated spend; (3) Media Review Gates (`AG08`) ensure generated visuals are checked before queuing; and (4) The Stop / Kill Switch (`AG08`) immediately halts further strategy-bound automated work while preserving already confirmed history.
3. Multi-Workspace Isolation and Team Roles for Brands & Agencies
When managing multiple brands—such as an outdoor equipment label (`Cycling Escapes`) and a historical documentary channel (`Świadek Dziejów`)—every brand must live in its own isolated workspace (`WS-01`, `WS-03`). Channels, strategies, Brand Voice profiles, conversations, and team permissions (`WS-02`) stay strictly scoped to each brand.
Comparison of the Three MUMRO Autonomy Modes
| Autonomy Mode | What AI Agents Can Do | Where Human Approval Sits |
|---|---|---|
| `draft_only` | Read context, analyze cohorts, and save article/post drafts | Humans review drafts and manually initiate all scheduling |
| `schedule_with_confirmation` | Prepare drafts + propose calendar slots via `prepare_*` | Human clicks Confirm on each prepared schedule action |
| `autonomous_schedule` | Schedule within approved strategy, allowed channels & cadence | Human sets strategy boundaries, quiet hours, credit budget & stop switch |
MCP Simulator: Prepare → Confirm Gate & 3 Graduated Autonomy Modes
Agent reads active Strategy v2, Brand Voice boundaries, quiet hours (22:00–07:00), and available schedule slots without mutating any state.
Agent saves a visible draft inside the FIELDWORK workspace. Creating a draft never triggers an external social API call.
MCP server validates autonomy rules and returns an exact human-readable preview (Instagram · Oct 11 · 18:00) plus short-lived token `cfm_9f82a1b4`.
After explicit human sign-off, the confirmation token commits the exact post to the MUMRO publishing plan with a full audit trail.
Configuring Autonomy Rules, Quiet Hours (22:00–07:00), and Workspace Roles
An agency team operates both workspaces simultaneously and needs clear boundaries on who can draft, who can approve strategy, and when AI agents can schedule.
- Set `schedule_with_confirmation` mode, a 1,000-credit monthly budget, and `22:00–07:00` quiet hours.
- Enabled mandatory human review for media and assigned Owner, Admin, Member, and Viewer roles.
- Verified that engaging the Stop switch immediately halts delegated actions without losing workspace drafts.
Workflow outcome: Complete peace of mind: AI accelerates the heavy lifting while humans own the rules.
You set the rules: 3 autonomy modes, quiet hours, budgets, and isolated workspaces.
MUMRO puts human control at the center of AI operations—with interactive autonomy controls, credit budgets, quiet hours, instant stop switches, and isolated multi-brand workspaces.
- Switch between `draft_only`, `schedule_with_confirmation`, and `autonomous_schedule` (`AG05`).
- Enforce allowed channels, cadence limits, quiet hours (`22:00–07:00`), monthly credit budgets, media review, and an instant stop switch (`AG06`–`AG08`).
- Run separate workspaces for every brand with Owner, Admin, Member, and Viewer roles (`WS-01`–`WS-03`).