How to check your Claude usage

To see exactly where you stand, open Settings > Usage in the Claude web or desktop app, or run /usage in Claude Code. Both show the same picture: a progress bar for each window, a percentage from 0 to 100, and the time that window next resets. Everything else on this page is how to read that screen and what to do about what it says.

Settings > Usage in the web and desktop apps

In claude.ai and in Claude Desktop, your usage lives under Settings > Usage. You get one row per window that Claude meters you on:

• A progress bar for each window, so you can see the shape of your consumption at a glance.
• A percentage from 0 to 100 for how much of that window you have used.
• The next reset time for each window individually.

You will see the rolling five-hour session window alongside the weekly caps, and the weekly side is tracked as two separate ceilings: one that applies across all models, and one that applies to Opus only. That is why you can be blocked from Opus while ordinary work continues fine.

Checking usage from Claude Code

You do not have to leave the terminal. One command covers it:

• /usage shows the same windows, percentages, and reset times you see in the app, plus a breakdown of what has been eating your allowance: skills, subagents, plugins, and individual MCP servers, with a day or week toggle. It runs immediately and does not interrupt a response, so you can check mid-task without killing whatever Claude is doing.
• The session block at the top of /usage shows token counts and a dollar figure computed locally at list rates. On a subscription that number is informational; it is the one that matters when you are billed by API key.
• /cost is an alias for /usage; they are the same screen. /usage-credits opens the usage-credits settings for your plan.

One caveat on the breakdown: the attribution figures are approximate and computed from session history on the machine you run the command on, so work you did on another device or in the browser is not reflected there. The plan usage bars themselves do cover your whole account.

The split matters: if you are on Pro or Max, the plan usage bars are your screen. If you are on API-key billing there are no usage windows to read at all, only a metered bill. See whether a Claude Pro subscription includes API credits for why those are two separate products.

Every surface counts against the same pool

A common misread: people check the web app, see plenty of room, and are surprised when Claude Code stops. Usage from claude.ai, Claude Desktop, and Claude Code all counts against the same limits. The number on your usage screen already includes your terminal work.

So a heavy agentic afternoon in the repo and a long browser conversation compound against one ceiling. When you are planning a big run, read the usage screen first rather than estimating from what you remember doing.

One reason a long session eats more than its activity suggests: Claude Code resends the whole conversation with every request. Prompt caching keeps that cheap, but the first message after a break longer than the cache lifetime misses the cache and reprocesses the full context. Clearing between unrelated tasks is the simplest fix.

Reading the percentages and when to act

The percentage is simply how much of that window you have burned, and Claude will warn you as a window gets close. Treat the warnings as pacing signals rather than alarms:

• Session window high, weekly low. You are burning fast in a short burst. Wait for the five-hour window to roll off; it refills automatically and you are usually counting in hours, not days.
• Weekly high. This is the one to plan around. Ship the work that matters, defer long multi-file agentic runs, and check the exact reset time on the screen before you rearrange your week.
• Opus cap high, general cap fine. Move routine work to a lighter model and save Opus for the parts that need it.

If a banner has already appeared, what the approaching weekly limit warning means covers the specifics. For reset timing, see when Claude limits reset: weekly caps reset at a fixed time assigned to your account, which is why there is no single published reset hour for everyone.

What "increase my weekly limit" actually means

There is no setting that raises your cap. The request resolves to one of three real actions:

• Upgrade the tier. Moving from Pro to Max, or up the Max tiers, raises the ceilings. This is the only change that lifts the caps themselves.
• Turn on usage credits. Paid plans can opt in to usage credits (Settings > Usage, or /usage-credits in Claude Code) and keep working past the cap. They are off unless you enable them, and they bill at standard API pricing rates, so this is real extra spend rather than a bigger allowance.
• Shift work off the subscription. Run the heaviest jobs on API-key billing instead, which is metered and has no weekly window. That is a different product with a different bill, not a bigger allowance.

And the honest default: wait. Both windows refill on their own with no action needed.

One rule worth adopting: your own usage screen beats any published quota, including this page. Anthropic changes limits, runs promotions, and adjusts plans; every specific message count you find in a forum thread is a snapshot from a week that has already passed. The in-app number is the only figure that reflects your account today.

If your usage screen is mostly empty

Most people who go looking for the usage screen find the opposite of what they feared: a weekly bar sitting well under half, week after week. That is worth noticing, because weekly capacity is a ceiling, not a balance. Whatever you have not used when the window resets is simply gone.

Tokens for Good exists to capture that recurring surplus. Your Claude Code claims a queued nonprofit, researches its real-world impact against a fixed methodology with citations, and submits a structured report. Every organization is researched twice by independent contributors, then validated, consolidated, scored deterministically, and human-reviewed before it reaches the public directory. It runs on the subscription you already pay for with no separate API cost, and it can run on a schedule in the background. To be clear, "tokens" means AI model tokens, not crypto: no coin, no wallet, no blockchain. See how the research works.

Frequently asked questions