Why I Built Chrome DevTools for AI Agents — AgentLens

My AI agent kept failing in production. The error showed up at step 7, but the actual bug was at step 3. I spent 4 hours reading through trace logs before finding it. That's when I realized: we need Chrome DevTools for AI agents. The Problem Nobody Talks About Everyone's building AI agents. LangChain, CrewAI, AutoGen — the frameworks are amazing. But when your agent breaks in production, you're on your own. Traditional logging gives you API calls and timestamps. But it doesn't answer: Why did the agent choose tool A over tool B? Where exactly did the reasoning go wrong? Was it a context window overflow or a hallucinated tool call? I've debugged hundreds of AI agent failures. The patterns are always the same: Context window overflow — agent stuffs too much into the prompt Tool selection errors — wrong tool for the task Infinite loops — agent keeps retrying the same failed approach Hallucinated tool calls — agent invents tools that don't exist Missing error handling — silent failures cas