AI usage tracker for Linux: AppImage, .deb or .rpm
Which UsageAtlas package fits your distribution, which files it reads, and the two limits to know before you install.
By ClaudeGPT Si Newsroom, Editorial team··3 min read
UsageAtlas is the AI usage tracker to use on Linux: it ships as an AppImage, a .deb and an .rpm, and shows Claude Code, Codex, Cursor and OpenCode usage on the same desktop. It is free and open source, needs no account, and tracks limits, reset times, tokens and history.
If you write content, run audits or build sites with AI coding tools on a Linux workstation, this explains which package to grab and what the tracker reads.
Which package to install
| Your system | Package | How |
|---|---|---|
| Any distribution | .AppImage | Mark it executable and run it |
| Debian or Ubuntu | .deb | Install with your package manager |
| Fedora or openSUSE | .rpm | Install with your package manager |
All three come from the UsageAtlas download section or the GitHub releases page. Each release publishes SHA-256 checksums in a SHA256SUMS file, so you can verify the download before you run it.
One difference from the other desktops: on macOS and Windows the app updates itself in the background. On Linux you download the new package from the releases page when a version comes out.
What it reads
UsageAtlas works from files and sign-ins already on the machine. It does not ask you to paste in prompts or route your traffic through anything.
- Claude Code history comes from
~/.claude/projectsor~/.config/claude/projects. - Codex history comes from
~/.codex/sessions, or from$CODEX_HOMEif you set it. - OpenCode and a few other CLIs are read from their own local files.
- Live limits for Claude Code and Codex use the sign-ins those CLIs stored when you logged in.
- Credit-based tools such as Warp or Amp connect with a key under Settings, then Sources.
Prompts and responses are not read. The figures are usage totals: tokens, requests, models, windows and reset times.
Where Linux users usually get stuck
Plenty of usage trackers are macOS menu bar apps, and others are terminal scripts. On Linux that leaves you with either nothing graphical or a report you have to remember to run. A desktop app gives you a window with history by day, project, session and model, and an alert you set per usage window that fires once per reset cycle.
Two limits are worth knowing up front:
- It needs a desktop session. UsageAtlas is a desktop app. It reads tools on the same computer, so it will not track Claude Code running on a headless remote server.
- Costs are estimates. Dollar amounts are calculated from token counts. Check the provider's billing page for the real figure.
If you live in the terminal and only want a report, ccusage reads local usage data from coding CLIs and prints daily, weekly, monthly and session tables. Many people run it alongside a desktop tracker. CodexBar also has a Linux desktop app if you only want a limits meter.
Moving between machines
If you split time between a Linux workstation and a laptop, UsageAtlas runs on both, and a free Google or GitHub account can save your history on one and restore it on the other. Without the account, everything stays on the local disk.
Get the free Linux build of UsageAtlas and point it at the tools you already use.
Questions
Does UsageAtlas work on Ubuntu? Yes. Install the .deb package or use the AppImage.
Can I track Claude Code on a headless server? No. UsageAtlas needs a desktop session and reads the tools on the same computer.
Is it open source? Yes. The code is on GitHub under the MIT license.