Agent configuration
The Reflex agent is configured primarily through the dashboard. Settings are synced to your server automatically — you rarely need to touch anything on the server itself.
Dashboard settings
The following settings are managed per-server in the Reflex dashboard under Server Settings.
Monitoring
- Metrics collection interval — How often the agent reports CPU, memory, disk, and process metrics
- Health check interval — How often the agent runs service health checks
- Log forwarding — Enable forwarding of system and application logs to the Reflex timeline
Playbooks
Enable or disable individual autonomous playbooks per server under Server → Playbooks. Team policy still applies: Settings → Repair playbooks (apply_mode: dry-run only vs auto-apply). See Repair playbooks.
Deployments
- Git repository — The repository to deploy from
- Branch — The branch to track
- Build commands — Commands to run during the build phase
- Release commands — Commands to run after a successful build (migrations, cache clearing, etc.)
- Health check URL — Endpoint to verify after deployment
Log levels
The agent log level can be changed temporarily for debugging via the dashboard. Available levels:
| Level | Description |
|---|---|
error | Critical failures only |
warn | Warnings and errors |
info | Normal operational messages (default) |
debug | Verbose output for troubleshooting |
Environment variables
| Variable | Description |
|---|---|
REFLEX_API_URL | Control plane URL (set automatically during installation) |
REFLEX_TEAM_TOKEN | Team registration token (found in dashboard Settings) |
REFLEX_WORKER_SYSTEMCTL | Queue worker systemctl lifecycle. Default on when PID 1 is systemd. Set to 0 to opt out on dev/sandbox hosts. |
Secrets
Treat REFLEX_TEAM_TOKEN (and any other credential env vars) as secrets:
- Never commit them to git or paste them into shared tickets/logs
- If you store them in a file (for example a systemd
EnvironmentFile), restrict mode:chmod 600 - Prefer values injected by the installer or systemd credentials over hand-edited world-readable env files
- If a token may have leaked, rotate it from dashboard Settings (same place it was issued) and reinstall/re-register the agent with the new value