Ask HN：你们是如何监控生产环境中的 AI 智能体的？
4 分•作者: jairooh•6 个月前
最近发生的几起事件（DataTalks 数据库被 Claude Code 擦除，Replit 代理在代码冻结期间删除数据），清楚地表明，在没有可观测性的情况下在生产环境中运行 AI 代理是危险的。
我见过的常见故障模式包括：无法逐步了解代理做了什么，未追踪的 token 使用导致意外的 LLM 账单，未被发现的风险输出，以及事后分析缺乏审计追踪。
我一直在构建 AgentShield (https://useagentshield.com) — 一个用于 AI 代理的可观测性 SDK。它提供执行追踪、输出风险检测、每个代理/模型的成本跟踪，以及对高风险操作的人工审核。通过两行代码集成到 LangChain、CrewAI 和 OpenAI Agents SDK 中。
好奇大家都在用什么。自己构建监控系统？LangSmith？Langfuse？还是只是听天由命？
查看原文
With the recent incidents (DataTalks database wipe by Claude Code, Replit agent deleting data during code freeze), it's clear that running AI agents in production without observability is risky.<p>Common failure modes I've seen: no visibility into what the agent did step-by-step, surprise LLM bills from untracked token usage, risky outputs going undetected, and no audit trail for post-mortems.<p>I've been building AgentShield (https://useagentshield.com) — an observability SDK for AI agents. It does execution tracing, risk detection on outputs, cost tracking per agent/model, and human-in-the-loop approval for high-risk actions. Plugs into LangChain, CrewAI, and OpenAI Agents SDK with a 2-line integration.<p>Curious what others are using. Rolling your own monitoring? LangSmith? Langfuse? Or just hoping for the best?