Understanding Agent Status Codes
A complete reference of every status code and live badge your agent reports in the dashboard and mobile app.
In this article you'll learn about Status code overview, then explore Online, Busy, and Idle mechanics, and finally review Error states & automated self-healing.
Status code overview
Every Botonom agent continuously broadcasts its operational status through live card badges in the web dashboard and mobile app:
| Code | Badge Color | Meaning | Token Usage |
|---|---|---|---|
ONLINE | 🟢 Green | Healthy, connected, and ready to accept new tasks | No |
BUSY | 🔵 Blue | Actively processing a task, reasoning, or executing tool calls | Yes |
IDLE | ⚪ Slate | Healthy and standby, with no inbound tasks for 15+ minutes | No |
PAUSED | 🟡 Amber | Manually paused by an administrator; queues incoming requests | No |
ERROR | 🔴 Red | Critical failure encountered; waiting for resolution or manual triage | No |
RECOVERING | 🟠 Orange | Platform watchdog is automatically restarting the container | No |
Status transitions are synchronized in real-time over WebSocket and reconciled with the local MMKV cache on mobile devices.
Online, Busy, and Idle mechanics
BUSY indicates the agent is executing prompt inference, querying skills (e.g. database, spreadsheets), or synthesizing documents.
ONLINE means the agent is listening on channels and ready for immediate interaction. If no inbound messages or scheduled tasks arrive for 15 minutes, the agent transitions to IDLE to conserve host resources.
Idle agents consume zero tokens and wake up in sub-second latency the moment a new channel event or user prompt is dispatched. You are never billed for idle time.
You can also programmatically inspect any agent's health via our REST API:
curl -H "Authorization: Bearer $API_KEY" \
https://api.botonom.com/v1/agents/<agent-slug>/status
Response:
{
"agent_slug": "executive-secretary",
"status": "ONLINE",
"last_active": "2026-09-11T12:45:00Z",
"uptime_percent": 99.9
}
Error states & automated self-healing
RECOVERING indicates an unexpected runtime crash or unhandled upstream exception occurred. The platform container watchdog detects interruptions within 5 seconds and restarts the agent process with all in-flight queue items safely preserved.
PAUSED is an administrative hold. Incoming channel inquiries and tasks are held in queue until you click Resume.
ERROR activates if an agent fails three consecutive auto-recovery cycles, triggering immediate notifications to workspace administrators.