Ask HN: How are you monitoring AI agents in production?

[flagged]

The dashcam analogy is sharp. I'd extend it: most tools record what happened (tool X was called, output was Y), but not why the agent deviated from the plan. That's the gap that actually hurts during post-mortems. In my experience, the useful question isn't "what did the agent do?" — it's "at step T, the agent's stated intent was Z, but it executed W instead. Was that a model drift, a context window issue, or a tool failure?" Without causal structure in the log, you're left correlating timestamps and guessing. The DataTalks/Replit incidents both had this signature: the deviation was visible in hindsight from the logs, but no system caught the intent-execution gap in real time.

The distinction between reversible and irreversible actions mentioned here is crucial, but there's an organizational layer to this problem that most monitoring tools miss entirely.When you scale past a single team, you inevitably end up with a fragmented stack. Team A builds a support bot in LangGraph, Team B builds a research agent in CrewAI, and Team C writes raw Python against the Anthropic API. If you rely on framework-level monitoring or prompt-level guardrails, your audit trail is completely fractured. You can't confidently tell a compliance officer what your synthetic workforce is doing.We realized that observability and governance cannot live inside the agent framework. They have to live in an independent execution layer that sits between the agents and your business systems. The agent proposes an intent, but the execution layer acts as the system of record—verifying authority, checking budget, and logging the action—before the API call is ever allowed to hit your database.

[flagged]