v1.2.3 - TUI + desktop app

Track AI coding token usage and costs locally.

Token Use reads Claude Code, Codex, Cursor, GitHub Copilot, and Gemini history already on your machine, stores it in a local archive, and breaks it down by cost, model, project, and tool, with optional quota gauges when you choose to sync them.

> Install Token UseLatest release
THE PROMISE

Your tokens never leave your machine. Neither do you.

01
No API keys
tokenuse never authenticates with any provider. It only reads files you already have.
02
No proxy
No traffic flows through tokenuse. Your editor talks to your model exactly the way it always has.
03
No telemetry
There is no analytics endpoint. Nothing about your usage is measured or sent anywhere.
04
Opt-in network only
Currency rates, pricing books, quota sync, and update checks all happen when you ask, not behind your back.
THE DASHBOARD

A local dashboard that answers the only question that matters:where did it go?

The TUI and desktop app share the same archive, model registry, pricing, and report flow. The desktop is an eight-screen sidebar app with dedicated pages per tool, a Coach page that grades how you work with 27 rules evaluated entirely on your machine, and Scrollback full-text search across every archived transcript; the terminal keeps it tight with Overview, Deep Dive, Usage, Coach, and Scrollback. A read-only MCP server can serve the same numbers to Claude Code or any MCP client, over stdio or a loopback endpoint. In the terminal, press 1-5 for time ranges, o/d/u to switch pages, / for Scrollback, and e for reports.

Desktop gallery

Eight sidebar screens, project and MCP detail, and five tool pages

Token Use desktop app Overview screen showing cost, calls, sessions, cache hit rate, input and output totals, current limits grouped by tool, an activity chart, and top project and top model tables.DesktopOverviewToken Use desktop app Analytics screen showing spend over time, daily spend stacked by tool, an hour-by-weekday activity heatmap, provider and tool share donuts, cache efficiency, and ranked usage tables.DesktopAnalyticsToken Use desktop app Coach screen showing an overall B+ grade, rules clean and findings tiles, prompt quality, session hygiene, code review, and tool mastery scores with trend sparklines, priority findings, and flow and pace panels.DesktopCoachToken Use desktop app Scrollback screen showing a search query with per-session grouped results, role-tagged and term-highlighted snippets, and match counts, filtered by tool and project.DesktopScrollbackToken Use desktop app Tools screen showing rolling 24-hour consoles for Claude Code, Codex, Copilot, Cursor, and Gemini with cost, calls, tokens, plan-limit gauges, and top models.DesktopToolsToken Use desktop app Models screen showing a provider-grouped model catalog with 24 hours, 7 days, 30 days, this month, and all time columns and an expandable per-tool breakdown.DesktopModelsToken Use desktop app Projects screen showing a full project index with cost, average per session, sessions, calls, last active, and tool mix for every project in the period.DesktopProjectsToken Use desktop app project detail page showing KPI tiles for cost, calls, sessions, cache hit, and AI code output, an activity pulse, a per-tool split donut, top models, By Activity task categories, and the full session list.DesktopProject detailToken Use desktop app Config screen showing display currency, pricing and exchange-rate downloads, archive actions, limit sync, the Data Sources diagnostics panel, the sample data switch, and desktop settings.DesktopConfigToken Use desktop app Config MCP Server panel showing the enable toggle, the loopback endpoint and port, a masked bearer token with reveal and copy buttons, and a copyable claude mcp add command.DesktopMCP ServerToken Use desktop app Claude Code page showing the 24-hour console with per-model windows, plus top projects, top models, and top sessions.DesktopClaude CodeToken Use desktop app Cursor page showing the 24-hour console with model usage, top projects, top models, and top sessions.DesktopCursorToken Use desktop app Codex page showing plan-limit gauges for five-hour and weekly windows, Spark model windows, model usage, top projects, and sessions.DesktopCodexToken Use desktop app Copilot page showing the AI Credits monthly gauge with used and remaining credits, model usage, top projects, and sessions.DesktopGitHub CopilotToken Use desktop app Gemini page showing the 24-hour console with Gemini 2.5 Pro and Flash usage, top projects, top models, and top sessions.DesktopGemini

TUI gallery

Five tabs plus session and config

Token Use TUI overview showing cost, calls, sessions, cache hit, input and output tokens, activity pulse, project spend, model spend, shell commands, and MCP servers.TUIOverviewToken Use TUI deep dive screen showing activity trend, by-project spend, top sessions, project spend by tool, model efficiency, shell commands, and MCP servers.TUIDeep DiveToken Use TUI usage screen showing Claude Code, Codex, Gemini, Cursor, and Copilot console usage panels with pulse charts, costs, calls, tokens, quota reset times, and plan names.TUIUsageToken Use TUI Coach screen showing an overall practice grade, per-group scores, triggered findings with occurrence lines, and an advisory setup panel.TUICoachToken Use TUI Scrollback screen showing a search input with per-session grouped results, role-tagged and term-highlighted snippets, and match counts.TUIScrollbackToken Use TUI session picker open over Deep Dive, listing recent sessions by date, project, cost, and calls with a type-to-search filter.TUISessionToken Use TUI configuration screen showing currency override, exchange rates, pricing books, limit sidecars, subscription quota sync, local files, and keyboard shortcuts.TUIConfig
SUPPORTED TOOLS

Five local parsers, one archive.

Token Use reads the session files your tools already write to disk. It normalizes local history into one browsable archive, while limits-only quota sync for Copilot, Claude.ai, and ChatGPT (Codex) stays separate and opt-in from Config.

Claude Code
~/.claude/projects/ + Claude Desktop
  • Exact tokens and cache reads
  • Bash commands and tool calls
  • MCP server and tool usage
liveparser docs ->
Codex
~/.codex/sessions/**/rollout-*.jsonl
  • Per-turn token_count events
  • Cached input plus reasoning output
  • Tool call ledger
liveparser docs ->
Cursor
Cursor state.vscdb
  • Exact tokens when present
  • chars / 4 estimate fallback
  • Agent KV rows
liveparser docs ->
GitHub Copilot
~/.copilot/ SQLite stores + VS Code Chat
  • Exact CLI usage when recorded
  • VS Code Chat transcript estimates
  • AI Credits quota gauges
liveparser docs ->
Gemini
~/.gemini/tmp/<project_hash>/chats/
  • Exact usage and cache reads
  • Thought token tracking
  • Tool calls
liveparser docs ->
GET STARTED

Install the current release.

Use Homebrew on macOS, or pull checksummed TUI and desktop builds from GitHub Releases for Linux and Windows.

$ brew install russmckendrick/tap/tokenuse
$ tokenuse

$ brew install --cask russmckendrick/tap/tokenuse-desktop
$ open -a "Token Use"
$ curl -L -o tokenuse https://github.com/russmckendrick/tokenuse/releases/latest/download/tokenuse-linux-amd64
$ chmod +x tokenuse
$ sudo install -m 0755 tokenuse /usr/local/bin/tokenuse
$ tokenuse
PS> Invoke-WebRequest -Uri "https://github.com/russmckendrick/tokenuse/releases/latest/download/tokenuse-windows-amd64.exe" -OutFile "tokenuse.exe"
PS> .\tokenuse.exe
TUI: macOS - Linux - Windows-Desktop: macOS - Linux - Windows-v1.2.3