How to track Claude Code and Codex usage in one place

UsageAtlas, a free, open-source desktop app released under the MIT license with builds for Windows, macOS and Linux, combines Claude Code and Codex usage into a single dashboard showing running totals of tokens, requests, models and estimated cost. The app reads Claude credentials already stored on the machine and talks to the signed-in Codex CLI to pull each tool's usage windows, plan and credit details, without reading prompts or responses and without requiring an account. UsageAtlas adds the combined view and history that Claude Code's /usage and Codex's /status commands do not keep, while cost figures remain estimates that users should verify against their Anthropic or OpenAI accounts.

To track Claude Code and Codex usage in one place, install UsageAtlas, connect both tools, and leave the tool filter off: you get one running total of tokens, requests, models and estimated cost, plus each tool's own usage windows and reset times. It is free, open source and runs on your desktop, so nothing needs a new account to get started. Why one total matters when you use both Plenty of developers split their day. Claude Code takes the refactor, Codex takes the test fixes, and a third agent picks up the rest. Each tool reports on itself only. Claude Code's /usage knows nothing about your Codex sessions, and Codex's /status knows nothing about Claude. The day that surprised you is usually the sum of both, and neither command can show that sum. UsageAtlas keeps both sources in the same dashboard. When Claude Code and Codex are both connected, the combined total is the default view. Cursor, OpenCode, GitHub Copilot and other supported tools join the same total if you use them. Set it up in four steps 1. Download the app. Builds are listed on usageatlas.com https://usageatlas.com/ for Windows, macOS and Linux. The code is on GitHub under the MIT license. 2. Sign in to Claude Code as usual. UsageAtlas reads the Claude credentials already stored on your computer, so there is no extra key to paste. 3. Sign in to the Codex CLI. UsageAtlas talks to the signed-in Codex CLI on the same machine to read its windows, plan and credit details where OpenAI reports them. 4. Open the Today view. Both tools show up side by side, with remaining capacity and the local reset time for each supported window. No account is needed for any of this. A free account is optional and only saves your usage history so you can restore it on another computer. UsageAtlas does not read prompts or responses. What you see for each tool | | Claude Code | Codex | |---|---|---| | Usage windows | Five-hour and weekly | Session and weekly, where reported | | Plan details | Plan and extra usage | Plan and credits, where reported | | Local activity | Tokens, requests, models | Tokens, requests, models | | In-tool command | /usage | /status | Split the total when you need to The per-tool filter is the useful switch. Filter to Claude when a five-hour window is close to the edge and you want to know whether to keep going. Filter to Codex when its weekly allowance is the bigger risk. Turn the filter off when you want the honest number for the day or the month. History works the same way. Pick a range, filter by tool, and compare model mix, busy hours and weekdays. You can also set a remaining-capacity threshold per window and get one desktop alert per reset cycle, which beats discovering a limit mid-task. What it does not replace Keep using the built-in commands. /usage in Claude Code and /status in the Codex CLI are the quickest checks inside a running session, and the source guide says as much. UsageAtlas adds the combined picture and the history those commands do not keep. Cost figures in UsageAtlas are estimates. For an actual bill, use your Anthropic or OpenAI account. If you only want a terminal report, ccusage https://github.com/ryoppippi/ccusage also reads Claude Code and Codex logs and prints daily tables; it suits people who live in the shell and do not need live windows or alerts. Who it suits UsageAtlas fits anyone who pays for both an Anthropic plan and a ChatGPT plan and wants to know which one is doing the heavy lifting. It also fits people who switch tools when one window runs low and want to see, at a glance, which tool still has room. If you only ever use one agent, its own command may be all you need; the tracker earns its place once a second tool joins the day.