What It Is

The nerve centre of the agent fleet.

The Command Center is Zodyria I, the Chief-of-Staff sister — she runs the house 24/7. From the dashboard you deploy agents, watch live fleet status, schedule recurring tasks, review every approval request, and see the audit trail for every action taken.

  • Deploy agents with a job, a lane and a name
  • Track every task, every token, every cent of cost
  • Schedule the night shift — 39 automations on file
  • Guardians report to you where you already are: Discord
  • Three doors in: Admin · Team · Client — each sees what they should
  • Billing-ready, Stripe-ready, on an open stack
A — just do it B — do it, then report C — ask the human first
Capabilities

What the Command Center can do.

Every capability is available through both the dashboard UI and via Discord commands. You choose the interface.

Agent Deployment

Spin up a guardian with a named role, a set of skills, an autonomy tier, and a schedule. The Chief-of-Staff routes work to the right lane automatically. Deploy in one command; tear down in one line.

Interface: Dashboard panel · Discord slash command

Live Fleet Status

See which agents are working, waiting, or idle. Each guardian has a status light, a timestamp of last action, and a link to its current output. The scan line sweeps the dashboard — nothing goes stale unnoticed.

Refresh: Real-time via WebSocket · Discord live channel

Task Scheduling (Night Shift)

Cron-based scheduler for recurring work: daily reports, weekly market scans, hourly triage, midnight backups. Jobs run whether you are logged in or not. Each schedule logs its last run and next run.

Granularity: Every minute · Timezone: configurable per job

Approval Queue (Class C)

Any action that requires human approval (money, keys, deletes, public posts) lands in the queue. The agent proposes with evidence; you review and approve or reject. Every decision is timestamped and logged.

Notification: Discord DM · Dashboard banner · Email fallback

Audit Ledger

Every action — who proposed, who approved, what happened, when — is written to an immutable ledger. No edits, no deletions. The ledger is the source of truth for operations, billing, and compliance.

Format: Append-only log · Exportable as JSON or plain text

Cost Meter

See token usage, model cost, and API spend per guardian, per task, per day. Free-tier-first: the fleet prefers zero-cost models and only escalates to paid when the job needs it. You set the budget.

Granularity: Per-token · Alert threshold: configurable
Approval System

Three tiers. One rule: humans decide the big things.

Every guardian operates within a clear autonomy tier. These are not policy documents — they are enforced in code, at the agent level, with no override switch.

Class A — Auto-Execute

The agent reads, acts, and reports. Used for routine, reversible, read-only or low-impact work: fetching data, triaging email, running scans. Reports are logged and optionally posted to Discord.

  • Market data scans
  • Inbox triage and tagging
  • Scheduled health checks
  • Context file maintenance

Class B — Execute & Report

The agent acts autonomously but must immediately report — via Discord, dashboard, or both. Useful for work that is safe but worth watching. Reports are timestamped and cannot be suppressed.

  • Competitor recon briefs
  • Draft generation and review
  • Automation recipe execution
  • Routine backup operations

Class C — Propose & Wait

The agent cannot act. It prepares a proposal with all evidence and presents it in the approval queue. Only a human (or a human-delegated secondary) can approve or reject. Written in the agent's core prompt — structurally unbypassable.

  • Trades and withdrawals
  • API key or secret changes
  • Permanent deletions
  • Public posts or announcements
  • Billing or plan changes

Tier Override — Impossible

No agent can promote itself from C to A. No agent can approve its own pending action. The autonomy tier is set at deploy time and enforced by the router layer. The only override is a human editing the agent's configuration — logged and audited.

This is by design. The family proposes. You decide.

Getting Started

How to use the Command Center.

Via Dashboard

The web dashboard is the full interface. Deploy, monitor, schedule, approve — everything in one browser window. Works on desktop and tablet.

Open the portal → log in with your credentials → the fleet loads automatically. No configuration needed on first boot.

Via Discord

Every guardian has a dedicated channel. Send a plain-language command to the bot; it routes to the right agent. Reports land in the same channel as threaded replies.

The Discord control plane mirrors the dashboard — you see the same status, the same logs, the same approval queue.

"The Command Center is the single source of truth. Everything that happens in the fleet, happens through it — and nothing happens without a record."

— Zodyria I · Chief-of-Staff