How do I enable session logging in AgentDev?

You can enable logging by running the command `/agentdev:debug-enable`, which creates a configuration file in your project's `.claude` directory.

What tools can I use to analyze the debug output?

The logs are formatted in JSONL, which is specifically optimized for processing with the `jq` command-line utility or modern JSON viewers.

Where are the debug log files located?

All debug sessions are saved as JSONL files within the `claude-code-session-debug/` directory in your project root.

How do I change the amount of data captured?

You can adjust the 'level' field in `.claude/agentdev-debug.json` to 'minimal', 'standard', or 'verbose' depending on your debugging needs.

Is my sensitive data protected in debug logs?

Yes, Debug Mode automatically redacts sensitive patterns such as API keys, bearer tokens, and cloud credentials from the logs.

AgentDev Debug Mode

Name: AgentDev Debug Mode
Author: MadAppGang

byMadAppGang

•

211

分析と監視

Monitor and record detailed agent session events, tool invocations, and performance metrics for deep observability.

概要

AgentDev Debug Mode provides comprehensive observability into Claude Code agent behavior by recording every interaction to crash-resilient JSONL files. It captures a wide range of events—including tool invocations, skill activations, agent delegations, and phase transitions—allowing developers to troubleshoot complex workflows and optimize session performance. Designed with security in mind, it automatically redacts sensitive information like API keys and enforces restrictive file permissions, making it an essential tool for developers building and auditing advanced AI agent behaviors.

主な機能

Three configurable verbosity levels: minimal, standard, and verbose
Automatic redaction of sensitive data including API keys and tokens
Secure file handling with owner-only read/write permissions
211 GitHub stars
Crash-resilient JSONL event logging for detailed session history
Granular tracking of tool results, agent delegations, and proxy requests

ユースケース

Auditing agent interactions for security compliance and quality assurance
Analyzing and optimizing AI workflow performance and execution timing
Troubleshooting unexpected agent behavior or complex tool failures

概要

主な機能

Three configurable verbosity levels: minimal, standard, and verbose
Automatic redaction of sensitive data including API keys and tokens
Secure file handling with owner-only read/write permissions
211 GitHub stars
Crash-resilient JSONL event logging for detailed session history
Granular tracking of tool results, agent delegations, and proxy requests

ユースケース

Auditing agent interactions for security compliance and quality assurance
Analyzing and optimizing AI workflow performance and execution timing
Troubleshooting unexpected agent behavior or complex tool failures