How to check your Codex usage

If you’re in a Codex session and want to know where your tokens went, type /usage. It opens a token-activity menu without making you leave the terminal. That command landed in Codex CLI v0.140.0, so if /usage isn’t recognized you’re on an older build. Check and update first:

codex --version
npm install -g @openai/codex@latest   # or: brew upgrade --cask codex

Check usage inside the CLI

1. Open Codex in any project: codex.
2. Type /usage and press Enter. The official reference describes it as “View account token usage or use a rate-limit reset.”
3. Pick a view from the menu, or jump straight to one:

/usage daily        heatmap, GitHub-contributions style
/usage weekly       bar chart by week
/usage cumulative   running total

The menu shows lifetime tokens, your daily peak, current streak, and longest task, plus a 52-week activity graph (details here).

For the current session only, type /status. The docs describe it as “Display session configuration and token usage”: active model, approval policy, writable roots, and how many tokens this session has burned.

Check plan limits on the web

/usage shows token activity, but not the rate-limit windows that decide whether Codex will actually run. The 5-hour and weekly caps live on the account dashboard at chatgpt.com/codex/settings/usage. That’s the surface to open when you want the real remaining-quota number instead of an estimate.

Codex bills against token-aligned credits. That switch happened April 2, 2026, so the figures you see are credits mapped onto input, output, and cached tokens. The per-plan message and credit math changes often enough that we keep it on one living page. See /limits/ for current numbers by plan and model.

Show usage in the status line

There’s an opt-in status line that can print your remaining five-hour-limit and weekly-limit as percentages at the bottom of the terminal, refreshed as you work. The reset-time tokens are still a feature request, and the five-hour token was pulled on July 12, 2026 when OpenAI temporarily removed the 5-hour cap. Don’t be surprised if only the weekly figure shows up.

If the numbers look wrong

• /status shows a stale or off usage percentage. This is known behavior: the /status number is often inaccurate or stale on the first run and needs a second /status to settle, and the 5-hour window went missing from /status on Plus (issue #18742, issue #32840). Cross-check against the web dashboard.
• CLI says “usage limit reached” but the dashboard shows quota left. That’s a documented mismatch on OpenAI’s side, and there’s a fix page for it: /errors/usage-limit-reached-but-quota-remaining/.

Related

• /limits/: per-plan limits, credit rates, and window rules
• /errors/usage-limit-reached-but-quota-remaining/: phantom limit fix
• /errors/usage-limit-draining-too-fast/: when credits vanish faster than expected
• /models/: model list and reasoning tiers