Agent Health & Monitoring

How to verify your agents are running, what they log, and what happens when AI is unavailable.

How Agents Run

Agents execute on fixed schedules as background jobs:

Agent	Schedule
Maintenance Agent	Daily batch check, every morning
GM Agent	Daily report, every evening
Utilization Agent	Weekly analysis, Mondays
Pricing Agent	Weekly review, Mondays
Customer Service Agent	On demand, 24/7

There's no start/stop button to manage — scheduling is handled by the platform.

Verifying Agents Are Working

The simplest health check is their output:

• GM Agent: a fresh daily report on your Reports page each morning
• Maintenance Agent: fleet alerts reflect current hours and thresholds
• Utilization / Pricing Agents: weekly analyses land on Mondays
• Customer Service Agent: the support chat responds

Missing output for more than a cycle? Contact support — agent infrastructure is monitored on our side too.

Audit Trail

Every agent action is written to your organization's audit log: which agent, what it produced, and when. All agent output is structured data (not free text), which keeps the log precise and reviewable — useful for compliance and for understanding why a recommendation was made.

Learning Loop Health

Agents get smarter through the learning loops. You can check loop health directly:

• GET /api/v1/learning/status — event counts per loop: pending, processed, failed
• A growing "pending" count with no "processed" movement is worth a support ticket
• POST /api/v1/learning/process — manually trigger aggregation if you want fresh insights immediately

Graceful Fallback

All five agents are built with graceful degradation: if the AI service is temporarily unavailable, agents fall back to deterministic behavior (threshold-based checks, template responses, queued messages) rather than failing. Your operations don't stop because an API hiccupped.

System Health

Platform-level health (API, database, cache) is visible at the API health endpoint. See System Status.

Coming Soon

• In-app agent health dashboard (last run, next run, per-agent status)
• Performance metrics per agent (approval rates, revenue impact, escalation rates)
• Configurable health alerts

Troubleshooting

Daily report didn't arrive?

• Check the Reports page directly — email delivery and report generation are separate
• One missed cycle can be transient; two in a row, contact support

Agent output looks stale?

• Weekly agents refresh Mondays; daily agents each morning — check the timestamp on the report
• Learning insights update as feedback processes; see Learning Loop Health above

Worried an agent did something wrong?

• Agents recommend rather than act on your rates and bookings — check the audit log to see exactly what was produced and when

See Common Issues for more.