Enables comprehensive session logging and event tracking to analyze agent behavior and optimize AI workflows.
AgentDev Debug Mode is a specialized diagnostic tool for Claude Code that captures every aspect of an agent session—including tool calls, skill activations, and sub-agent delegations—into crash-resilient JSONL files. By providing granular visibility into the inner workings of complex AI workflows across multiple detail levels, it allows developers to troubleshoot errors, audit performance, and refine agent instructions with precision while ensuring sensitive data like API keys remains automatically redacted.
主な機能
01Detailed event logging of tool invocations, agent delegations, and phase transitions
02Automated sensitive data protection that redacts API keys, tokens, and passwords
033 GitHub stars
04Standardized event schema with correlation IDs for tracing related operations
05Crash-resilient JSONL output format optimized for streaming and CLI processing
06Configurable debug levels (minimal, standard, verbose) to balance detail and storage
ユースケース
01Troubleshooting unexpected agent behavior or failed tool executions in real-time
02Analyzing and optimizing the performance and duration of multi-step AI workflows
03Auditing AI interactions for security compliance and quality assurance purposes
