A local-first macOS dashboard for Codex CLI token usage, cost estimates, model usage, and project usage from your local ~/.codex logs.
Download the latest release · 中文 README
Use Codex Usage Desktop if you run Codex CLI on your machine and want an OpenAI Codex usage tracker that helps you review recent token usage, cost trends, cache behavior, model usage, and project usage without uploading local logs.
- Understand how many Codex tokens you use across recent work.
- Estimate Codex CLI costs by model, day, and project.
- See cache hit rate, cost per million tokens, and daily usage trends.
- Keep usage analysis local by reading
~/.codexsession logs on your machine. - Export the selected dashboard window to Excel or Markdown for reporting.
- Local-first scanning of Codex CLI session logs.
- Dashboard windows from 1 day to 90 days.
- Monthly usage totals for longer-term review.
- Token, cache, cost, model, and project breakdowns.
- Excel (
.xlsx) and Markdown (.md) exports. - Local SQLite cache reset and rebuild when usage data needs to be refreshed from scratch.
Download the latest macOS build from GitHub Releases.
Current release builds:
- macOS Apple Silicon
- macOS Intel
Windows and Linux builds are planned, but the current release workflow only publishes macOS desktop packages.
- Open the desktop app.
- On first launch, the app reads your local Codex session logs from
~/.codexand builds a local usage cache. - Use the Dashboard view to see total tokens, estimated cost, daily averages, cache hit rate, cost per million tokens, daily trends, model usage, and project usage.
- Switch the dashboard window between 1d / 2d / 7d / 14d / 30d / 60d / 90d.
- Open the Monthly view to see natural-month usage totals.
- Click
Rescan local logsafter new Codex sessions to refresh the dashboard. - Use
Exportto save the selected dashboard window as Excel (.xlsx) or Markdown (.md). - Use Settings ->
Reset cacheif the local cache needs to be cleared and rebuilt from your Codex logs.
- Source Codex logs are read locally and are not deleted by scan or reset actions.
- Usage summaries are stored in a local SQLite cache in the app data directory.
- Pricing data is cached locally. If the pricing cache is missing, the app tries to fetch Codex model pricing from LiteLLM; if that fails, it uses the bundled pricing table.
- Cost values are estimates based on the pricing data available to the app.
CODEX_HOME: Codex home directory, default~/.codexCODEX_USAGE_TIMEZONE: timezone for day bucketing, default system timezone with UTC fallback
- The app shows daily and monthly aggregate usage, not session-level detail.
- Unknown models default to zero cost.
- Release packages are currently macOS-only.
- Windows and Linux release packages.
- More detailed usage drilldowns.
- More export and reporting options.
The app is built with React 19, Vite, Tauri v2, and a native Rust usage pipeline.
Install Node.js >= 24, pnpm, Rust, and Tauri v2 system dependencies, then run:
pnpm install
pnpm dev:appUseful checks:
pnpm test
pnpm typecheck
cd src-tauri && cargo testPackaged builds use:
pnpm build
pnpm tauri build