Troubleshooting
Agent health: degraded
Degraded means an agent ran but could not complete its full task. Offline means an agent did not run at all. Both states appear in the portal dashboard and trigger an Operator escalation card.
Health states
Online
The agent ran successfully on its last scheduled run. All MCP connections responded normally.
Degraded
The agent ran but produced incomplete output. Usually caused by a partial MCP failure: the agent could read from some sources but not all.
Offline
The agent did not run at its last scheduled time. The Operator will have created an escalation card with the reason.
Step 1: Read the Operator escalation card
When an agent shows as degraded or offline, the Operator creates an escalation card in your approval inbox. Open the card first. It will say exactly which MCP source failed, what error it returned, and the steps to fix it.
Follow the instructions in the card before continuing here. The card is specific to your situation; this guide covers the general patterns.
Step 2: Identify the cause
MCP connection failure
Most degraded states are caused by a dropped MCP connection. Runclaude mcp listand look for any source showing an error. Use the relevant reconnect guide:
Schedule not firing
Run claude routines list to check whether the agent schedule is registered. If missing, re-run the agent's first-run setup to restore it.
Handbook missing or invalid
Agents refuse to run without a valid handbook. Check the portal for handbook errors. If the handbook is invalid, complete the handbook interview again to produce a valid one.
Step 3: Trigger a test run
After fixing the underlying issue, trigger a manual run from the portal event log (Dashboard → Event log → Run [agent name] now). If the run completes and the agent produces output, the issue is resolved. The health indicator in the dashboard updates within a minute.
If the issue persists
Check the event log for error events from the affected agent. Each error event includes the full error message. If the error message is not covered by the reconnect guides above, note the exact error text and contact support with it.
The event log is always available at app.flostackos.com/dashboard even when agents are offline.