Monte Carlo’s Agent Health flags stale LLM in its own troubleshooting bot
Monte Carlo’s internal Agent Health monitor caught its Troubleshooting Agent calling an outdated Anthropic Haiku LLM in 595 daily calls. The silent drift could have wasted credits and degraded diagnostics, showing that continuous trace‑analysis is essential to keep production agents up‑to‑date and reliable.