AgentLens

AI agents fail in production in ways that traditional logs can't catch. They hallucinate, enter infinite loops, or corrupt tool state. You have no idea why it happened until you spend hours digging through flat text logs. AgentLens is a "flight recorder" for your autonomous AI agents. It captures full execution hierarchies—every LLM call, tool execution, and reasoning trace—and streams them to a live dashboard. Features: 2-Line Integration: Add it to your existing OpenAI/Anthropic setup without changing your code. Trace Hierarchy: See nested spans and tool calls, not just flat logs. Real-Time Dashboard: Debug your agent while it runs via live WebSocket updates. Open Source & Self-Hostable: MIT licensed, built on a lean stack (FastAPI/Postgres), deployable in minutes.