How to check your Claude Code and Codex usage limits

You're halfway through a refactor when the agent stops: limit reached, back in three hours. Each tool shows how close you are, if you know where to look:

• Claude Code: type /usage. It shows the current 5-hour session, the current week, and a weekly limit for one model, each as "% used" with the time it resets. Subscription plans only.
• Codex: type /status. The card shows your limits, such as the 5-hour and weekly ones, as "% left" with the reset time. You need to be signed in with ChatGPT.
• Grok Build: type /usage. The Usage limit tab shows your weekly or monthly limit as a percentage used and the reset date.
• Keep it on screen: Claude Code and Codex can show the limits in their status line all the time.
• The catch: each tool shows only its own limits, from inside a session. Codex counts what's left; Claude Code and Grok count what's used.

How do I check my Claude Code usage limit?

Inside a session, type:

/usage

Claude Code opens its settings on the Usage tab. You'll see a bar for each window, with the share used and when it resets, in your local time:

• Current session: the 5-hour window.
• Current week (all models): the 7-day window.
• Current week for one model, when your plan has a separate limit for it.

Below that is Usage credits. With credits on, Claude Code keeps working after you hit a limit instead of stopping; /usage-credits turns them on or asks your admin.

The numbers are fetched when you open the tab, so they're current. /usage also answers to /cost and /stats. If you use an API key instead of a subscription, it says usage limits are only available for subscription plans. /status shows your account and model, not your limits.

How do I check my Codex usage limit?

Inside a session, type:

/status

Codex prints a card with the model, folder, and account, followed by one row per limit, for example:

Weekly limit:  [████████████████████] 99% left (resets 21:18 on 9 Oct)

Rows appear only for the windows your plan has: a 5-hour limit, a weekly one, sometimes daily or monthly, and separate rows for models with their own limit. A Credits row shows up if you have credits. Codex counts down: "99% left" means you've used 1%.

The numbers come with Codex's responses. If they're more than 15 minutes old, /status says the limits may be stale; send a message to refresh them. The card also links to chatgpt.com/codex/settings/usage for the account view.

/usage is a different command in Codex. It charts your recent token use rather than your limits.

How do I check my Grok Build usage limit?

Inside a session, type:

/usage

A window opens with three tabs: Context usage, Usage limit, and Session info. Press Tab to switch. The Usage limit tab shows:

• Weekly limit or Monthly limit, depending on your billing period.
• A bar with the percentage used.
• Resets: the date and time it starts over.
• Any prepaid credits, then this session's token use.

Grok fetches the numbers when you open the window. On a team plan it says usage limits are managed by your team. /usage also answers to /cost.

From the shell, grok usage <session-id> prints one session's tokens and cost as JSON. It doesn't show how much of your limit is left.

How do I keep my usage visible while I work?

Claude Code passes rate limits to your status line command. Add the command to ~/.claude/settings.json:

{
  "statusLine": { "type": "command", "command": "~/.claude/statusline.sh" }
}

Save this as ~/.claude/statusline.sh. It reads the JSON with jq; if jq --version says the command isn't found, install it first, for example with brew install jq:

#!/bin/sh
input=$(cat)
five=$(echo "$input" | jq -r '.rate_limits.five_hour.used_percentage // empty')
week=$(echo "$input" | jq -r '.rate_limits.seven_day.used_percentage // empty')
[ -n "$five" ] && printf '5h %.0f%%' "$five"
[ -n "$week" ] && printf '  7d %.0f%%' "$week"

Then make it executable with chmod +x ~/.claude/statusline.sh.

The limits are included for Pro and Max subscribers, or behind a gateway that sets a spend limit; on other plans the line stays empty. They come from your last response, so they appear after the first message and update as you go. Or type /statusline and describe what you want; Claude Code writes the script for you.

Codex has /statusline to pick items, or add them in ~/.codex/config.toml:

[tui]
status_line = ["five-hour-limit", "weekly-limit"]

Each item is hidden when your plan doesn't have that window.

Grok Build's status line script gets token and cost totals, but not your limit. Use /usage for that.

Why are my limits different in each tool?

They're separate plans with separate rules. Claude Code has a 5-hour window and a weekly one, plus a weekly limit for some models. Codex shows whichever windows your plan has: a 5-hour one, a weekly one, and on some plans daily or monthly. Grok Build counts by week or by month, depending on how you're billed.

They also count in different directions. Claude Code and Grok show how much you've used. Codex shows how much is left. "30%" in Claude Code is close to "70% left" in Codex.

Hitting one tool's limit doesn't touch the others. When Claude Code runs out mid-task, you can carry on in Codex or Grok Build; see How to hand off a task from Claude Code to Codex.

Can I see all three limits in one place?

Yes. In PonyMux, click the link button at the bottom of the sidebar to open Agent Integrations. Each connected agent has a card with its usage bars: the share used and how long until each window resets. The numbers refresh every three minutes while the panel is open.

PonyMux's Agent Integrations panel: a card each for Claude Code, Codex, and Grok, with usage bars and reset times.

Each agent's numbers come from where that agent keeps them:

• Claude Code: PonyMux sends one tiny request with your Claude Code sign-in and reads the 5-hour and 7-day usage it reports back.
• Codex: the same usage service Codex uses, with its 5-hour, daily, and weekly windows.
• Grok Build: the last billing entry in Grok's own log, so it's as fresh as your last Grok turn.

You connect each agent from the same panel. For what else the connection adds, see Agent Integrations.

---

Originally published on ponymux.com, where the demos are interactive.