How to check your Claude Code usage limits, and when they reset

Type /usage in Claude Code. On a Pro or Max plan it shows how much of your current five-hour session and of your weekly limit you have used, each with the day and time it resets. The same limits are on claude.ai under Settings > Usage, and Claude Code hands them to your status line after every reply, so you can keep them on screen while you work.

Last updated: October 2026

Written by the developer of Crest, 8 October 2026. Run that day on a 16-inch MacBook Pro (Mac16,7) on macOS 27.0.1 with Claude Code 2.1.292 on a Claude subscription; the percentages shown here are made up, the rest is what Claude Code printed. Anthropic’s pages are listed under Sources. Crest comes last, and you do not need it.

Where to look

All of them read the same two limits, and the limits are shared: Anthropic says they “are shared across Claude and Claude Code, meaning all activity in both tools counts against the same limits.” A long chat on claude.ai moves the numbers Claude Code shows, and the other way round.

What /usage prints

On Claude Code 2.1.292 the top of /usage looks like this (the percentages are made up; the lines and the reset times are as our run printed them):

Current session: 23% used · resets Oct 9 at 2am (Asia/Beirut)
Current week (all models): 41% used · resets Oct 12 at 11pm (Asia/Beirut)
Current week (Fable): 8% used · resets Oct 12 at 11pm (Asia/Beirut)

What's contributing to your limits usage?

The third line is there because the plan has a separate weekly limit for Fable, which Anthropic says Max plans have. Under it, “What’s contributing” splits the last 24 hours and the last 7 days by habit, such as sessions that ran for 8 hours or more and requests over 150k tokens of context, with the subagents and MCP servers that used the most. Anthropic calls those figures approximate and says they come from the sessions on this Mac only, so work on another Mac or on claude.ai is not in them.

It also works outside a session: claude -p "/usage" printed the same report in our terminal and exited, which is handy in a script. On an API key there is no plan limit; the Session block of /usage shows the token cost instead.

When the limits reset

Anthropic’s help center puts it simply: a session limit “that resets every five hours”, and weekly limits “that reset once every week”. The exact moment is the one /usage prints next to each line, in your own time zone. On our account the session reset at 2am and the week on Monday at 11pm, Beirut time, so do not assume midnight.

Before you run out, Claude Code warns you in the session with a line like You’ve used 85% of your session limit · resets 3:45pm. At the limit it stops with You’ve hit your session limit or You’ve hit your weekly limit, with the reset time. Switching models does not get around either of those; it does get around a model limit such as You’ve hit your Opus limit. From there you can wait, buy usage credits with /usage-credits, or, on Claude Code 2.1.234 or later, let the session wait and carry on by itself after the reset.

Keep them on screen: a status line

Claude Code runs your status line script after every reply and passes it a JSON object. On Pro and Max that object has a rate_limits part with five_hour and seven_day, each with used_percentage and resets_at in Unix seconds. Save this as ~/.claude/statusline.sh:

#!/bin/sh
# ~/.claude/statusline.sh: the model, then both usage limits with their resets.
input=$(cat)
model=$(printf '%s' "$input" | jq -r '.model.display_name // "Claude"')
five=$(printf '%s' "$input" | jq -r '.rate_limits.five_hour.used_percentage // empty')
week=$(printf '%s' "$input" | jq -r '.rate_limits.seven_day.used_percentage // empty')
five_at=$(printf '%s' "$input" | jq -r '.rate_limits.five_hour.resets_at // empty')
week_at=$(printf '%s' "$input" | jq -r '.rate_limits.seven_day.resets_at // empty')
out="$model"
[ -n "$five" ] && out="$out | 5h $(printf '%.0f' "$five")% (resets $(date -r "$five_at" '+%H:%M'))"
[ -n "$week" ] && out="$out | 7d $(printf '%.0f' "$week")% (resets $(date -r "$week_at" '+%a %H:%M'))"
printf '%s\n' "$out"

Then point Claude Code at it in ~/.claude/settings.json (if the file already has other keys, add statusLine next to them):

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

We fed the script three payloads shaped like the one Claude Code 2.1.292 writes on this Mac: both limits, only the weekly one, and none. It printed:

Opus | 5h 23% (resets 00:24) | 7d 41% (resets Mon 09:44)
Sonnet | 7d 88% (resets Fri 23:24)
Opus

Three things that explain an empty status line. rate_limits only arrives after the first reply of a session, only for Pro and Max subscribers, and a window drops out once its reset time has passed, which is why the script prints whatever is there and skips the rest. date -r is the macOS form; on Linux use date -d @"$five_at".

Or on the notch, with a warning before you run out

Crest, a Mac notch app, puts both limits on its Agents screen, next to your sessions: “Claude limits” with the 5 hours share and the time it resets, and the 7 days share, in red from 85%. It reads the same rate_limits through a small status line bridge it installs, which runs the status line you already had, so yours keeps working, and puts it back exactly as it was if you turn the feature off.

• A warning before, not after. When the 7-day limit crosses 85%, or when the 5-hour window would run out before it resets at the pace of your last half hour, a warning shows on the closed notch, such as “About 40 min left at this pace”. The pace one is Crest’s estimate and says so. Each comes once per window. The switch is Settings > Coding agents > “Warn before limits run out”, on by default.
• Never a stale number. If the last reading is more than 30 minutes old, or the window has reset, the row is not drawn at all.
• Codex too. Crest shows Codex’s limits under Claude’s, from Codex’s own session files; how to check Codex usage limits has the details.
• What is free: answering Claude Code, Codex and Copilot CLI prompts from the notch is free, with no account. The limits on the Agents screen and the warnings are part of Crest Pro.

1. Download Crest. Signed and notarized.Free tier, no account · 7-day Pro trial, no card · macOS 14 or later
2. Open Crest’s Settings, go to Coding agents and turn on “Allow or deny commands here”. That installs Crest’s hooks and the status line bridge in one go.
3. Run a Claude Code session; the limits appear after its first reply.

Crest Pro, with Take and Minutes, is $24.99 once or $2.99 a month, and signing in gives you 7 days of it free. More on the notch side: Claude Code in the notch.

Sources

• Anthropic, Claude Code docs, Manage costs effectively. Read 8 October 2026. /usage, its plan usage breakdown on Pro, Max, Team and Enterprise, and that those figures are “approximate and computed from local session history on this machine”.
• Anthropic, Claude Code docs, Customize your status line. Read 8 October 2026. The rate_limits.five_hour and rate_limits.seven_day fields, that they appear only for Pro and Max subscribers and after the first reply, that a window drops once it resets, and when the status line runs.
• Anthropic, Claude Code docs, Errors. Read 8 October 2026. The session and weekly limit messages, the 85% warning, which limits switching models gets around, and the desktop app’s usage ring.
• Anthropic Help Center, Use Claude Code with your Pro or Max plan and Usage limit best practices. Read 8 October 2026. The five-hour session and weekly limits, that they are shared with Claude, the separate Fable limit on Max, and Settings > Usage.
• Our own run, 8 October 2026, on a 16-inch MacBook Pro with macOS 27.0.1 and Claude Code 2.1.292: claude -p "/usage" (its lines above, with our percentages swapped for made-up ones), the rate_limits object as Claude Code wrote it for the status line on this Mac, and the script above fed three payloads.
• Crest source, read 8 October 2026: UsageBridge.swift, AgentSessionsReading.swift, QuotaWatch.swift, SettingsCatalog.swift, and the 6.1.0 entry of Changelog.swift.

FAQ

Keep reading: Codex usage limits · Copilot CLI notifications · Claude Code notifications · Codex notifications · Claude Code vs Codex · Crest overview