Date Compiled: 2026-04-28
Type: source
Related Questions: factory-methodology, autonomous-cron-management, multi-agent-operations, ai-fleet-management

SuperAda: Enterprise Operations — 136 Active Autonomous Tasks

The Scale

At any given time, the Enterprise Crew runs:
- ~98 active autonomous cron jobs (surface-level count: 60 standalone + 38 inside cluster groups)
- 9 Lobster pipelines (typed JSON envelope + resumable approvals workflow)
- 10+ n8n workflows (webhooks, system integrations, third-party API syncing)
- 19 localized watchdog crons on individual agent gateways

True total: 136 autonomous background tasks across the fleet.

Cluster Architecture

Cron clusters group related tasks to save compute and context window overhead:

Cluster Schedule Tasks Inside
cluster:hourly-ops Hourly 8 sub-tasks
cluster:morning-intelligence Daily 7:00 7 sub-tasks
cluster:2x-daily-intel Twice Daily 5 sub-tasks
cluster:morning-data-pipeline Daily 06:05 5 sub-tasks
cluster:morning-actions Daily 8:00 4 sub-tasks
overnight-proactive-work Daily 3:00 3 sub-tasks
cluster:6h-maintenance Every 6h 2 sub-tasks
evening-data-collection Twice Daily 2 sub-tasks
research-build Daily 6:00 2 sub-tasks

The Four Meta-Crons

When operating at this scale, things break. Four meta-crons manage the fleet.

1. The Model Orchestrator (The Fixer) — Every 6h

Acts as dynamic load balancer and crisis responder.

Behavior:
- Health Check: Runs check-providers.sh → writes provider-status.json
- Audit: Lists all enabled crons, auto-fixes deprecated model references
- Recovery: On rate_limit/429/All models failed → check provider health → re-run
- Delivery Fixing: Detects "thread not found" → updates delivery to #mail-room fallback

Key: Fixes problems immediately. Does not log and queue — fixes now.

2. The Cron Watcher (The Architect) — Every 6h

Unified cron governance engine.

Behavior:
- Lobster Detection: Reads payloads. 3+ deterministic sequential steps → tags LOBSTER_CANDIDATE
- Clustering: Groups small crons with same schedule into cluster crons
- Auto-Implementation: Takes JSON snapshot, applies safe fixes, logs rollback manifest

Safety: Only safe operations are auto-implemented (batch delivery migration, Lobster conversion, model fixes, cluster merges).

3. The Cron Health Report (The Watchdog) — Twice Daily (8am & 8pm)

Passive scan of the full fleet.

Flags:
- lastStatus = error → 🔴 FAILING
- lastDurationMs > 5 minutes → 🟡 SLOW (potential infinite loop)
- Should have run but no lastRunAtMs → 🟡 NEVER-RAN
- Model name typos → 🟡 TYPO

Posts emoji-coded report to Discord #medbay.

4. The Auto-Lobster Converter — Sundays 2am

Dry-run validator. Scans LOBSTER_CANDIDATE crons, runs lobster-converter-core.sh, posts summary to #upgrades. Does not modify crons — validates for human review.

n8n External Watchdog Pattern

n8n (self-hosted workflow engine) is monitored externally by OpenClaw meta-crons.

cluster:hourly-ops includes an n8n-health step: pings n8n API externally. If n8n goes down, Ada detects it and alerts in Discord #upgrades or #medbay before Henry notices.

Principle: Don't run watchers inside the system being watched.

Lobster Pipelines

Lobster is the crew's typed JSON envelope + resumable human-in-the-loop approvals workflow runtime. Complex multi-step crons are being migrated to Lobster pipelines.

Active pipelines:
- hourly-ops.lobster
- email-triage.lobster
- entity-backup.lobster
- daily-ops.lobster
- secondary-inbox-monitor.lobster
- morning-actions.lobster
- evening-data-collection.lobster
- morning-data-pipeline.lobster
- production-deploy.lobster

Agent-Localized Background Jobs

Each agent runs localized tasks on their own independent gateway, not counted in the primary 98.

Spock (100.78.229.38)

  • spock-research-digest (08:00, 20:00)
  • crewlink-spock-post (every 4h)
  • local-session-prune (04:30)

Scotty (Raspberry Pi 100.68.207.75)

  • github-auto-sync (every 6h)
  • self-improvement-daily/weekly
  • watercooler-scotty
  • mc120-daily-standup-automation
  • Vehicle Market Scanner

Zora (Mac Studio 100.86.150.96)

  • Zora Hourly Work Cycle — massive continuous loop: knowledge management, goal tracking, opportunity scouting, vault curation
  • ada-stability-monitor
  • superada-daily-tts (local TTS podcast)
  • zora-ada-sync
  • Weekly Vault Health + Goal Review
  • moneyz-ceo & Blog Managers (6 watchdog/publishing jobs)
  • local-session-prune

Key Operational Insights

  1. Cluster before you scale — Group same-schedule tasks to save compute/context overhead
  2. Watchdogs must be external — Don't monitor your monitoring system from inside itself
  3. Auto-fix, don't just log — The Model Orchestrator fixes deprecated models, delivery targets, and re-runs immediately
  4. Rollback manifests before changes — Snapshot before auto-implementation so fixes are reversible
  5. Cron governance compounds — The Cron Watcher makes the fleet progressively more efficient each run
  6. Health reports are passive — They're for issues the Orchestrator isn't authorized to auto-fix

Source

  • https://superada.ai/blog/the-98-active-crons
  • https://superada.ai/blog/managing-68-autonomous-crons