Agent installed but doesn't appear in Fleet
Most common causes:
- Service didn't start. Check
Get-Service diagnosly-agent, should beRunning. IfStopped, runStart-Service diagnosly-agentas admin. - Token already used. Install tokens are single-use. Generate a new one and re-run register.
- Network blocked. The agent must reach the gateway over HTTPS. Check Windows Firewall + corporate proxy.
- Agent on the wrong tenant. If you have multiple
agent.envprofiles, use--profile <name>explicitly to avoid alphabetical auto-pick.
Command hangs in running forever
Either the agent crashed mid-execution, or the script blocks on user input. Background timeout job will mark it timeout after the per-type threshold (see Command audit trail). Force-cancel via the Cancel button on the command row.
AI diagnose fails with "Claude Code not found"
The agent locates an installed Claude Code automatically. If it isn't found, install Claude Code via the dashboard's Deploy AI action, that's the recommended path for production.
If diagnose fails with "API key not found", visit Settings → LLM Providers and confirm the active provider has a valid key. Demo tenants don't ship with a key, bring your own Anthropic / OpenAI / GLM key.
Notifications
The bell in the topbar shows unread notifications. Types:
alert_triggered · alert_escalated · device_offline · device_online · task_completed · task_failed · script_failed · system
Alerts are generated when device metrics exceed configured thresholds (CPU, memory, disk). Alert escalation runs every 60 s, unacknowledged alerts escalate to critical after a configurable timeout.
:::callout type="info" Heartbeat error log shows "Failed to send heartbeat: connect refused", the agent and gateway can't talk. Verify the gateway is reachable from the target machine and that no proxy/firewall is in the way. :::