skills.rest
Olatisunkanmi
OlatisunkanmiCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

cm-get-token-insights

Parse JSONL conversation logs into per-turn token and cost analytics.

8|1|Updated Mar 14, 2026
Search Tags:#analytics#claude#dashboard#cost-optimization#claude-code#token-insights#workflow-profiling
One-click install
npx skills add https://github.com/Olatisunkanmi/claudefiles --skill cm-get-token-insights
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cm-get-token-insights
Source: https://github.com/Olatisunkanmi/claudefiles/tree/main/skills/cm-get-token-insights
Command: npx skills add https://github.com/Olatisunkanmi/claudefiles --skill cm-get-token-insights

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Many Claude Code users lack visibility into where tokens are consumed and why costs rise; this Skill turns raw conversation logs into per-turn analytics, cost breakdowns, and actionable recommendations so teams can reduce waste and optimize workflows.

Core Features & Use Cases

  • Ingest & Parse: Parse JSONL conversation files from local Claude project directories into per-turn analytics tables and session summaries.
  • Cost Optimization: Identify top waste drivers by dollar amount, recommend CLAUDE.md rules and model substitutions, and estimate savings.
  • Workflow Analytics & Dashboard: Produce an interactive offline HTML dashboard showing model economics, top projects, skill/agent/hook usage, hook latency, and week-on-week trends.
  • Agent Enrichment: Construct data-grounded prompts for a claude-code-guide agent and spawn it in the foreground to generate personalized guidance for the top insights.
  • Use Case: Run on your ~/.claude/projects to discover cache-miss hot spots, model-mix savings opportunities, and underused or failing skills that drive cost.

Quick Start

Run cm-ingest-token-data to parse your ~/.claude/projects and open the generated dashboard at ~/.claude-memory/dashboard.html for an immediate spending and workflow report.

Frequently Asked Questions about cm-get-token-insights

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I analyze Claude Code token usage from JSONL conversation logs?▼

You can analyze Claude Code token usage by parsing JSONL conversation logs from ~/.claude/projects directories into per-turn analytics that break down token consumption, cache hits, and cost drivers.

What is the best way to identify Claude API cost drivers and reduce waste?▼

Identifying Claude API cost drivers and reducing waste involves computing waste_usd-ranked insights from conversation logs to highlight cache-miss hot spots, model-mix savings opportunities, and underused skills.

Does this token analytics tool work with local Claude project directories?▼

Yes, this token analytics tool works with local Claude project directories by ingesting JSONL outputs from ~/.claude/projects to produce project-level and session-level summaries without requiring external API connections.

Can I generate an interactive dashboard for Claude Code workflow profiling?▼

You can generate an interactive offline HTML dashboard for Claude Code workflow profiling that displays model economics, top projects, skill and hook usage, hook latency, and week-on-week spending trends.

How do I get personalized recommendations for optimizing Claude model economics?▼

You get personalized recommendations for optimizing Claude model economics by spawning a claude-code-guide agent in the foreground with data-grounded prompts constructed from your parsed analytics to generate customized guidance for top insights.

What are the limitations of analyzing Claude conversation logs locally?▼

Analyzing Claude conversation logs locally is limited to parsing JSONL files physically present under ~/.claude/projects, meaning it cannot provide insights for remote sessions or historical logs that have been deleted or moved.

Related Skills

Explore More Software Engineering Skills→
tneztnez
3

channels-read

Read channel messages with filters and return chronological metadata.

Community
Intermediate
kennedym-dskennedym-ds
14

performance-analysis

Analyze software performance and output structured optimization recommendations.

Community
Advanced
jdeweedatajdeweedata

rules-organizer

Organize Claude Code instructions into modular rule files under .claude/rules.

Community
Basic
felvieirafelvieira
22

cost-tracker

Aggregate tokens, API usage, and execution time to generate per-skill and session cost reports.

Community
Advanced