Claude Code usage: check it, know your limits, read the cost

Type /usage in a Claude Code session; /cost and /stats open the same screen. On a Pro, Max or Ultra plan you see plan usage bars. On the API you see a dollar estimate.

Checked against Anthropic docs on November 12, 2026.

Check your usage in Claude Code

  • Session block. Input, output, cache-read and cache-write tokens by model, with a dollar total for the session. It is meant for API users, and it resets when /clear starts a new session.
  • Prompt cache line (Claude Code v2.1.251 or later). The share of input served from cache, the number of misses, and whether the cache is warm. From v2.1.260 it can also name the likely cause of the last miss.
  • Plan usage breakdown (Pro, Max, Ultra, Team, Enterprise). What counts against your plan, split by skills, subagents, plugins and MCP servers, with recommendations. Press d or w to switch between the last 24 hours and the last 7 days. From v2.1.305, press f to act on the recommendations: Claude Code shows you each change before making it.
  • These figures come from session history on this machine. Usage on other devices or on claude.ai is not included.
  • In VS Code, /usage opens the Account & usage dialog, with the recommendations and a Day/Week toggle.

Check usage outside the terminal

  • claude.ai, Settings > Usage: how much of the current session you have used, the time remaining, and your next reset time.
  • Claude Console: usage and spend for API keys.
  • Analytics dashboards: for Team and Enterprise admins and owners, and for Console users with usage access.

Plan limits: the 5-hour session and the weekly limit

  • Pro and Max: a session limit that starts with your first message and resets in full five hours later, plus weekly limits across all models.
  • Ultra: verified Ultra accounts have no 5-hour session limit.
  • Some plans also have a separate weekly limit for Fable.
  • Limits are shared across claude.ai, Claude Code and Claude Desktop: all of them count toward the same limit.

Anthropic does not list token amounts for these limits and says it may adjust them, so this page doesn’t give any. Settings > Usage shows yours and when it resets.

What the cost figure means

  • On Pro, Max and Ultra, Anthropic says the session cost figure “isn’t relevant for billing purposes.” Your plan covers the usage.
  • Claude Code works out the figure on your machine from token counts at list price. If your organization’s admin sets a modelPricing table in managed settings, it uses those rates instead.
  • For scale, Anthropic reports that across enterprise deployments the average is about $13 per developer per active day and $150 to $250 per developer per month, with costs below $30 per active day for 90% of users.
  • With usage credits on, /usage shows your spend for the month against any monthly limit you set. /usage-credits manages them.

Why your usage runs out faster than expected

  • Cache misses. Changing the system prompt or the tool set, or switching models, forces the cached context to be rebuilt. On older models, changing effort does too. The prompt cache line shows how often it happens. Editing CLAUDE.md mid-session does not break the cache, but the edit only applies from the next session.
  • Fresh input is the expensive part. A cache read costs a tenth of fresh input on most Claude models, a twentieth on Opus 5.5 and Sonnet 5.5, and a fortieth on Fable 5.1 and 5.2 (Anthropic list prices, November 12, 2026).
  • Context you carry every turn. A large CLAUDE.md, many MCP servers and long-running loops add tokens to every request. The plan usage breakdown names the biggest ones.
  • What helps: pick the model at the start of a session, /clear between unrelated tasks, /compact when a session gets long, and keep CLAUDE.md short.

Track usage across sessions

Claude Code keeps session transcripts in ~/.claude/projects/ for 30 days by default; the cleanupPeriodDays setting changes that. Tools that read them:

  • ccusage: a daily and monthly dollar ledger.
  • npx sigrank: input, output, cache-write and cache-read counts across the available time windows, with your Yield worked out on your machine.

More ccusage alternatives

Usage tells you how much. Yield tells you how well.

Two people can hit the same limit and get very different amounts of work out of it. SigRank scores that with Yield: cache read × output ÷ input². Fresh input is squared, so re-sending context drags the score down fast and reusing it pushes the score up. How Yield works · How to improve your Yield

Run npx sigrank to see yours. Nothing is published until you run sigrank submit, and sigrank submit --dry-run shows the exact payload first. Prompts, responses, file paths and repo or project names are never sent.

npx sigrankSee the board →

FAQ

How do I check my Claude Code usage?
Type /usage in a Claude Code session. For your session usage and reset time, open Settings > Usage on claude.ai.
Is /cost the same as /usage?
Yes. /cost opens the same screen, and /stats opens it on the Stats tab.
Is the /usage dollar amount my bill?
Not on a subscription. Anthropic says the figure isn't relevant for billing on Pro, Max and Ultra. On the API it is an estimate at list price.
What are the Claude Code limits on Pro and Max?
A session limit that resets in full five hours after your first message, plus weekly limits. Anthropic doesn't list token amounts and may adjust them. Settings > Usage shows yours.
When does my Claude Code limit reset?
The session limit resets five hours after the first message of the session. Settings > Usage on claude.ai shows your next reset time.
Why does Claude Code use so many tokens?
Usually cache misses and context carried on every turn: a large CLAUDE.md, many MCP servers or long loops. The plan usage breakdown in /usage shows which.
Does SigRank read my prompts?
No. Prompts, responses, file paths and repo or project names are never sent, and nothing is sent until you run sigrank submit.