The author built AgentLens, an open-source Python SDK for AI agent governance, after encountering a silent agent failure in a multi-agent pipeline where a misleading alert masked the real cause—a missing file. The SDK provides an audit trail for every LLM call, offering visibility into agent actions. It is designed to address the lack of observability in autonomous agent workflows.
dev.to