Show HN: AgentDbg - local-first debugger for AI agents (timeline, loops, etc.) (github.com)

AgentDbg has introduced a local-first debugger for AI agents, enabling developers to capture structured traces of agent executions, including LLM and tool calls, errors, and state updates. With just a few lines of code, users can enable tracing and visualize the entire execution timeline without needing cloud services, accounts, or telemetry. This tool is particularly significant for the AI/ML community as it addresses common challenges in debugging AI agents, such as silent loops, tool mismatches, and non-deterministic behavior—issues that conventional logging often fails to clarify. AgentDbg operates entirely on local machines, storing all trace data in easily inspectable JSON files. It provides a user-friendly interface that reveals critical metrics like run status, duration, LLM and tool call counts, along with detailed insights into errors and loop warnings. Designed to be framework-agnostic, it effortlessly integrates with any Python codebase. Future updates are expected to bring features like deterministic replay and enhanced integrations with existing AI frameworks, marking AgentDbg as a vital addition to the development toolkit for AI agents.