termdeck
Claude Code has one command for usage: /usage. /cost and /stats are aliases for it in Claude Code 2.1.309, so older guides that tell you to run /cost still work. What it shows depends on how you signed in: a claude.ai subscription has usage limits, an API key has a bill.
Everything here was checked on 23 October 2026 against Claude Code 2.1.309, Anthropic's cost documentation and the Claude Help Center. Anthropic does not publish fixed message counts per window, so this post does not either.
Which command shows what
| Command | What it shows |
|---|---|
/usage | Session cost, plan usage limits and activity stats in one screen |
/cost | Alias for /usage |
/stats | Alias for /usage that opens on the Stats tab |
/status | Version, model, account and connectivity: which plan and organization you are signed in under |
/context | How full the context window is, as a grid, with what is taking the space |
/usage-credits | Opens usage credits settings, or asks your admin for them, when you hit a limit |
/insights | An HTML report on how you work across recent sessions on this machine |
/usage runs immediately even while Claude is responding, so you can check it mid-turn.
Checking usage on a Pro, Max, Team or Enterprise plan
Run /usage. Subscribers see plan usage bars: Current session for the 5-hour window, Current week (all models) for the weekly window, and on some plans a separate weekly bar for one model family. Each bar says when it resets.
Below the bars, on a subscription plan, /usage breaks down what counted against your limits:
- Attribution: the share of recent usage from skills, subagents, plugins and each MCP server.
- Recommendations: one row per change worth making, each with the share of your usage it applies to, for example trimming a large
CLAUDE.mdor lowering effort. - Loops: the heaviest
/loopor scheduled tasks that ran recently.
Press d or w to switch between the last 24 hours and the last 7 days. When the footer shows f to fix, press f and Claude Code applies the recommendations it can (version 2.1.305 or later).
Two caveats from Anthropic's docs. The breakdown is computed from local session history on this machine, so usage from your other computers or from claude.ai is not in it. And the Session block at the top shows a dollar figure that is not your bill: your plan includes the usage, so that number only matters on an API key.
The Help Center also mentions /status for remaining allocation. In 2.1.309, /status shows your account, plan and organization, and the usage bars are on the Usage tab of the same screen, which /usage opens directly.
How the 5-hour session and weekly limits work
Anthropic's Help Center article How do usage and length limits work? describes two windows on paid plans: a session limit that resets every five hours, and weekly limits that reset once a week. Using Claude Code with your Pro or Max plan adds that Max plans also have a separate weekly limit for Fable.
The same articles say how fast you use them depends on the length and complexity of the conversation, the model, the effort level, and tools and connectors, which they call token-intensive. In Claude Code that last point is most of it: every request carries your whole conversation, so one short question in a session that has been open all day still uses the usage of the whole context.
When you hit a limit, Claude Code says which one: "You've hit your session limit" or "You've hit your weekly limit", with the reset time. Those windows are shared across all models, so switching with /model does not get you back in. A model-specific limit, such as an Opus limit, is different: switching to another model family does. Your options are to wait, turn on usage credits with /usage-credits, or upgrade. /rate-limit-options lists them, and from 2.1.234 Claude Code can wait and continue the task automatically after the reset.
Does Claude Code use the same usage as claude.ai?
Yes. On Pro and Max, the Help Center says the limits are shared across Claude and Claude Code: a long chat on claude.ai or in the desktop app uses the same allowance as your coding session. On Team and Enterprise, Anthropic's docs say each member's Claude Code usage draws from a per-seat allowance with the same 5-hour and weekly windows, shared with Claude chat and Cowork.
This is also why /usage on one machine can look low while you are near a limit: its breakdown counts only this machine, but the bars come from your account.
Checking Claude Code cost with an API key
With an API key or a Claude Console login there are no plan limits, only a per-token bill. /usage (or /cost) shows the Session block:
Total cost: $0.55
Total duration (API): 6m 20s
Total duration (wall): 6h 33m 10s
Total code changes: 0 lines added, 0 lines removed
Usage by model:
claude-sonnet-4-6: 1.2k input, 5.3k output, 940.0k cache read, 50.0k cache write ($0.55)That example is from Anthropic's docs. Claude Code works the figure out locally from token counts at list price, and it resets to $0 when /clear starts a new session. It is an estimate. The authoritative number is the Usage page in the Claude Console. The first time you sign Claude Code in with a Console account, it creates a workspace called "Claude Code", so its spend is already separated there. An admin can set a workspace spend limit on it, and the Claude Code page in the Console shows spend per team member.
For a sense of scale, Anthropic's docs put the average across enterprise deployments at around $13 per developer per active day and $150 to $250 per developer per month, with 90% of users below $30 per active day. Your number depends mostly on the model and how long you let sessions grow.
If you have a subscription but /usage shows a growing dollar figure and no limit bars, an API key is probably overriding your login and you are paying per token. See Claude Code ignores your subscription login.
Keep usage visible in the status line
To watch usage without running a command, add a status line script. Claude Code passes it JSON that includes cost.total_cost_usd and, for Pro and Max subscribers, rate_limits.five_hour.used_percentage and rate_limits.seven_day.used_percentage with their resets_at times. The rate limit fields appear only after the first response in a session, so read them with a fallback:
jq -r '.rate_limits.five_hour.used_percentage // empty'Third-party Claude Code usage trackers
Claude Code keeps its transcripts as JSONL files under ~/.claude/projects, with token counts on each response, and several open source tools read them. They are not made by Anthropic, and they estimate cost from tokens just as /usage does:
- ccusage (
npx ccusage@latest) prints daily, weekly, monthly and per-session reports from local data, for Claude Code and several other agent CLIs. - Claude Code Usage Monitor is a live terminal monitor with predictions and warnings.
- claude-usage is a local dashboard for tokens, costs and session history.
All three see only the machine they run on, and none of them can see what you used on claude.ai. For your real remaining limit, the bars in /usage are the source.
How model and effort change consumption
The two settings that move usage most are the model and the effort level. Anthropic's advice is to use Sonnet for most coding work and keep Opus for hard architecture or multi-step reasoning. Thinking tokens count as output, so lowering effort with /effort on simple tasks cuts usage directly. Our guide to Claude Code models and effort covers which to pick and how to set them.
The other habits that matter, from the same docs: /clear between unrelated tasks, a short CLAUDE.md, MCP servers you are not using turned off in /mcp, and a smaller model for subagents. A session left open for hours, a /loop firing on an interval or a long break that lets the prompt cache expire can all use far more than the work looks like it should.
Usage across several machines in Termdeck
/usage and the trackers above answer for one computer. If you run Claude Code on a laptop, a desktop and a server, Termdeck puts the numbers in one place. Its Usage page scans the transcripts on every connected machine and shows API-equivalent cost, input and output tokens and sessions, with cost per day, the model mix and a table by project that drills down to single runs. It counts subagent transcripts, which are easy to miss, and only the totals leave each machine.
The sidebar also shows the live 5-hour and weekly windows for the Claude account signed in on a machine, and Settings shows them for every saved login side by side. That makes it easy to see an account is nearly out before you start a long run. Switching to a second account before a limit is covered in how to switch Claude accounts.
FAQ
How can I see my Claude Code usage?
Run /usage inside Claude Code. On a Pro, Max, Team or Enterprise plan it shows your 5-hour session and weekly limit bars with reset times, plus a breakdown of what used them. With an API key it shows the cost of the current session. /cost and /stats open the same screen.
Does Claude Code use my usage?
Yes. On Pro and Max, Claude Code and Claude share the same usage limits, so chatting on claude.ai and coding in Claude Code draw from one allowance. On Team and Enterprise, Claude Code draws from your seat allowance, which is shared with Claude chat.
What is the difference between /usage, /cost and /stats?
In Claude Code 2.1.309 there is no difference in what they show: /cost and /stats are aliases for /usage. /stats opens the same screen on the Stats tab.
How much does Claude Code cost?
On a subscription it costs your plan price, and usage is limited by the 5-hour and weekly windows rather than billed. With an API key you pay per token. Anthropic reports an average of around 13 dollars per developer per active day across enterprise deployments, with 90 percent of users under 30 dollars per active day.
When does the Claude Code usage limit reset?
The session limit resets every five hours and the weekly limit once a week. The exact reset time for each is shown on its bar in /usage and in the message you get when you hit it.
Is the cost in /usage what I will be billed?
No. Claude Code estimates it locally from token counts at list price. On a subscription your plan covers the usage, and on an API key the Usage page in the Claude Console is the authoritative figure.
Can I track Claude Code usage across several computers?
Not with /usage alone: its breakdown counts only the machine you run it on. Local tools such as ccusage have the same limit. Termdeck adds up usage from every machine you connect on one Usage page.