Can I track token usage and costs with this skill?

Yes, the skill tracks total tokens, input/output tokens, and specific metrics for prompt cache hits and creations to help you monitor LLM efficiency.

What does Judgeval Tracing do for Claude Code?

It automatically captures detailed logs of your Claude Code sessions, including every prompt, model response, and tool execution, sending them to Judgeval for analysis and debugging.

Does it support subagent tracking?

Yes, it creates nested spans for subagents, allowing you to visualize exactly how Claude Code delegates tasks and how those sub-tasks perform.

Where are the hook logs stored locally?

Logs for the tracing hooks are written to the local file system at ~/.claude/state/judgeval_hook.log for easy troubleshooting.

Judgeval Tracing

Name: Judgeval Tracing
Author: JudgmentLabs

byJudgmentLabs

0•

Analytics & Monitoring

Enables comprehensive observability for Claude Code by automatically tracing sessions, LLM calls, and tool usage to Judgeval.

Judgeval Tracing for Claude Code provides developers with deep visibility into their AI coding sessions by automatically capturing and logging every interaction. It meticulously tracks session lifecycles, individual turn prompts, model input/output spans, and tool invocations like file operations or terminal commands. By integrating directly into the Claude Code CLI via system hooks, it helps teams debug complex agentic workflows, monitor token usage—including prompt caching—and audit LLM behavior for improved reliability and performance.

Key Features

010 GitHub stars

02Detailed LLM span logging with input/output and token usage

03Support for nested subagent spans and complex workflows

04Tool invocation tracking for file operations and terminal commands

05Automatic session and turn lifecycle tracking

06Prompt cache metric monitoring for cost optimization

Use Cases

01Auditing AI agent tool usage and terminal execution history

02Debugging complex multi-turn coding sessions and agent logic

03Monitoring LLM cost and performance via granular token usage metrics

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add judgmentlabs/judgeval-claude-plugin trace-claude-code

For use in Claude.ai and ChatGPT

Download Skill