How to monitor an AI agent running in production?

Use structured logs + Telegram/Discord alerts + heartbeats + health check endpoints