How to check your Claude Code usage limits

The quick answer: /usage

In any Claude Code session, type:

/usage

On a subscription plan this shows your plan usage bars: how much of the current 5-hour session and the weekly limit you’ve used, and when each resets. Below that it breaks down what counted against your limits (skills, subagents, MCP servers, scheduled loops) with suggestions to cut usage. Press d or w to switch that breakdown between the last day and the last week.

/cost and /stats are aliases for the same screen. /usage runs immediately even while Claude is working, so you don’t have to wait for the turn to end.

The Session block at the top shows a dollar figure. On Pro, Max and Ultra that number is a list-price estimate for interest only; your subscription covers it.

The two windows

Window	How it works	Message when you hit it
Session (5 hours)	Starts with your first message and resets 5 hours later	“You’ve hit your session limit”
Weekly (7 days)	Caps total use across the week; some plans also have a per-model weekly cap	“You’ve hit your weekly limit” or “You’ve hit your Opus limit”

The session window is what most people run into during a busy afternoon. The weekly one bites heavy users of agents and subagents, because every subagent and parallel task draws on the same allowance.

On the web

The same numbers are on claude.ai under Settings › Usage (claude.ai/settings/usage). That page also shows your usage credits, if you’ve turned them on to keep working past the limit. From Claude Code, /usage-credits opens it.

When you hit the limit

Run /rate-limit-options. It offers to wait and continue the interrupted task automatically when the window resets, to use usage credits, or to upgrade. Recent versions open that menu on their own when a request is blocked.

Keep the limits in your status line

If you’d rather not type /usage, a status line script can show the percentages all the time. Claude Code passes a rate_limits object to the script for Pro, Max and Ultra subscribers, after the first response in a session.

Save this as ~/.claude/statusline.sh and make it executable with chmod +x:

#!/bin/bash
input=$(cat)
five=$(echo "$input" | jq -r '.rate_limits.five_hour.used_percentage // empty')
week=$(echo "$input" | jq -r '.rate_limits.seven_day.used_percentage // empty')
reset=$(echo "$input" | jq -r '.rate_limits.five_hour.resets_at // empty')
[ -n "$reset" ] && reset=$(date -r "$reset" +%H:%M)
[ -n "$five" ] && printf '5h %.0f%% (resets %s) · week %.0f%%' "$five" "$reset" "${week:-0}"

Then point Claude Code at it in ~/.claude/settings.json:

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

resets_at is Unix time in seconds, which is why the script formats it with date -r. The script needs jq (brew install jq).

Why usage climbs faster than expected

A few patterns account for most surprise limits:

• Long sessions. Every request carries the whole conversation. A one-line question in a session that’s been open all day still costs the full context. /clear between unrelated tasks.
• Cache misses after a break. Coming back after the prompt cache expires reprocesses everything.
• Subagents and loops. Each runs its own requests. /usage shows their share.
• Several sessions at once. Two terminals running agents burn the 5-hour window twice as fast.

The /usage breakdown lists which of these applies to you, as a share of recent usage.

Seeing it in the menu bar

Stat Space’s Usage limits panel puts the same numbers in the Agents tab and on the pinned Agents card: remaining percent per window (5-hour, weekly, per-model weekly), the reset time, and a pace line such as “limit in ~40m” or “on pace”. It’s free and off until you turn it on in Settings › Agents. For Claude Code it asks Anthropic’s usage endpoint at most every 5 minutes with the sign-in Claude Code already stored, keeps the token in memory only and never refreshes it, and only while the Agents tab or card is on screen.

Next to the limits, each running session shows its context use, tokens per minute and an estimated cost, so you can see which one is eating the window. More on the Agents feature.