The short answer
Type /usage inside Claude Code. On a Pro, Max or other claude.ai plan it shows how much of your current 5-hour session and your weekly limit you have used, and when each one resets. The same numbers are on claude.ai under Settings, then Usage. If you want them on screen all the time, with a warning before you hit the cap, a Claude Code usage tracker such as Edith's Usage extension keeps them in the menu bar.
How Claude Code limits work
A claude.ai subscription gives you two windows that matter for Claude Code. The session window covers 5 hours from your first message. The weekly window covers 7 days. Anthropic counts usage from claude.ai, the Claude desktop app and Claude Code against the same plan, so a long chat in the browser leaves less room for your terminal sessions. Anthropic does not publish fixed message counts per plan, which is why you check the percentage rather than counting prompts.
If you use Claude Code with an API key instead of a subscription, there is no plan limit: you pay per token, and the useful number is cost.
Method 1: /usage inside Claude Code
Built into Claude Code
Run /usage in any session
/usage shows the session cost, your plan usage limits and activity stats. On a paid plan it also breaks down what counted against the limit. /cost and /stats open the same screen, and Claude Code runs /usage right away even while Claude is still answering. The command list is in Anthropic's Claude Code commands reference.
/status is the other one worth knowing. It shows the Claude Code version, model, account and connectivity, which tells you whether you are signed in to the plan you think you are.
Method 2: claude.ai usage settings
Open claude.ai in a browser, then Settings, then Usage. The page lists the current session, the weekly limit and any per-model weekly window, each with its reset time. Because every Claude app draws from the same plan, this is the number to check when you split your day between the browser and the terminal.
Method 3: put the limits in the status line
Claude Code runs the command in the statusLine setting of ~/.claude/settings.json and passes it a JSON document. For Pro, Max and Ultra subscribers that document includes rate_limits.five_hour and rate_limits.seven_day, each with a used_percentage and a resets_at time, after the first response in a session. A one-line status line that prints the 5-hour figure looks like this:
jq -r '.rate_limits.five_hour.used_percentage // empty'Set it as the command of a statusLine entry with "type": "command". Anthropic's status line documentation lists every field.
Where the built-in methods fall short
- You have to ask.
/usageand the settings page answer when you look, not when you are about to run out. - The status line lives in one terminal. It only updates while a Claude Code session is open, and only after a response arrives.
- No history. None of them tells you which project or which day burned the week, or what the same work would have cost at API prices.
- One agent at a time. If you also run Codex or Cursor, their limits live somewhere else.
The Edith way: Usage
Usage polls Claude's usage API with the login Claude Code already saved on your Mac, so the 5-hour window, the weekly window and, when your plan reports one, the separate weekly Fable window show up as rings with a reset countdown. It reads that login from the Keychain without prompting and without changing it. It also reads the history Claude Code keeps on your Mac to show tokens and cost by day, model and repository. Your prompts and code are not read.
- 1
Install Usage
Find Usage on the Extensions page in Edith, or install it with one command.
- 2
Collect your history
Open Usage and press refresh. The first collection can take a few minutes on a Mac with a long Claude Code history.
- 3
Read your limits
Usage shows each window as a ring in the app and in the menu bar. From a terminal,
limitsprints the newest reading for every provider, and--refreshasks Edith to poll again first. - 4
Turn on limit alerts
In Usage settings, turn on alerts. Edith tells you when a window is almost capped, 90 percent by default, and when your recent pace would reach the cap before the window resets.
ed usage alertsshows what each window would alert about right now and why.
Every report also prints JSON, so a script can stop starting new work when the session window is nearly full:
ed usage limits --json | jq -r '.[] | select(.provider == "claude") | .session.percent'
ed usage summary --range today
ed usage models
ed usage projectsWhat you get beyond the limit
Because Usage keeps a history, you can see a heatmap of daily spend, which models did the work, and which repositories the tokens went to. The same view covers Codex, Cursor and other agents that keep usage on your Mac, so one menu bar readout covers all of them. If you run Claude Code on other machines over SSH, Usage can collect from those too with ed usage machines collect.
Pair it with the coding agent dashboard to see which agents are running while the limit climbs, and read how to check Codex usage limits if you split work between both. There is more on running Claude Code with Edith on the Claude Code on Mac page.
Common questions
Does Edith need my Anthropic password or an API key?
No. It uses the login Claude Code already saved. If Claude Code is signed in, Usage can read limits.
Why does /usage differ slightly from Edith?
Both read the same plan, but at different moments. Edith shows the time of its last reading in ed usage limits, and a refresh brings them back in line.
Can I see cost on a subscription?
Yes. Usage prices your token history at API rates, which is a useful way to compare days and projects even when your plan is a flat fee.