Skip to main content
skill CX Codex

get-token-insights — Claude Code skill: what it does, how to install

get-token-insights parses Claude Code conversation history, flags token waste, and generates an offline dashboard covering costs, hooks, and agent workflows.

no license ★ 276 updated 2026-10-29

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add gupsammy/Claudest
/plugin install claude-memory@Claudest
Open on GitHub

What get-token-insights does

The skill processes conversation logs in ~/.claude/projects/*/ through ingest_token_data.py. It calculates expenses across individual projects and models, tracks hook latency, and measures prompt cache misses. A dedicated claude-code-guide subagent then processes the telemetry to draft targeted cost reduction advice.

Output appears directly in the terminal and compiles into a standalone report at ~/.claude-memory/dashboard.html. When rapid five-minute cache invalidation drains tokens, the workflow can install protective scripts into ~/.claude/hooks/ after backing up settings.json.

Parsing around 2500 files during the initial pass takes roughly 100 seconds, while subsequent runs take under 5 seconds. The tool requires python3 in your environment.

What it helps with

  • uncover root causes of token overspending in Claude Code sessions
  • render an offline HTML dashboard tracking spend and agent metrics
  • spot slow or failing hooks across active configurations
  • evaluate financial losses triggered by frequent prompt cache invalidation

Example prompts for the agent

  • Analyze my token consumption and suggest cost reductions
  • Review hook runtimes and subagent delegation patterns from recent logs
  • Generate the spend dashboard in ~/.claude-memory/dashboard.html

How to install get-token-insights

1Plugin marketplace

/plugin marketplace add gupsammy/Claudest
/plugin install claude-memory@Claudest

This installs the whole claude-memory plugin with all its skills

2Skills CLI (Vercel)

npx skills add gupsammy/Claudest --skill get-token-insights -a claude-code

3Manually: put the files in place

~/.claude/skills/get-token-insights/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/get-token-insights/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about get-token-insights

How to install get-token-insights in Claude Code?

Add the repository and install the plugin using /plugin marketplace add gupsammy/Claudest, followed by /plugin install claude-memory@Claudest.

What does get-token-insights do?

It parses project conversation files, highlights model spending, assesses prompt caching efficiency, and builds an HTML analytics report.

What dependencies does this skill require?

It relies on a working python3 installation and read access to ~/.claude/projects/.

Similar skills

All skills for Agent setup: memory, context, orchestration →

From the same set47

The whole set →
recall-conversations recall-conversations reads back your earlier Claude sessions: finds a decision, restores context, or builds a retro from the local conversation database. CX Codex +1 ★ 276repo stars make-readme make-readme inspects repository context, runs a brief questionnaire, and builds a complete README.md with status badges. CX Codex +1 ★ 276repo stars push-pr push-pr is a Claude Code skill that pushes a branch and creates or updates its GitHub pull request. For large or mixed changes, it can propose splitting them into stacked PRs. CX Codex +1 ★ 276repo stars brainstorm brainstorm — an interview skill that turns the tables: the agent asks, you answer, and the session ends with a written spec or decision doc. CX Codex +1 ★ 276repo stars extract-frames extract-frames detects shot boundaries in a video and saves the first or last frame of each shot. Use it to review an edit or collect frames from selected shots. CX Codex +1 ★ 276repo stars search-youtube search-youtube finds YouTube videos, pulls transcripts and metadata, downloads audio and walks a channel's uploads through a single CLI the agent calls on its own. CX Codex +1 ★ 276repo stars improve-skill improve-skill audits skill effectiveness, simulates user requests, and pinpoints missing capabilities in instructions. CX Codex +1 ★ 276repo stars performance-auditor The performance-auditor subagent inspects code for algorithmic, data-access, and resource bottlenecks. Use it after changes on hot paths or when you need to reason about how code scales. CX Codex ★ 276repo stars

Project sponsors

Companies that keep this analytics open