Usage and limits

Claude Code usage and cost: what you pay, how to check it, how to use less

Updated November 17, 20268 min read

Claude Code usage is counted against your Claude plan: a 5-hour window and a weekly window on Pro and Max, and a weekly window only on Ultra for verified accounts. Type /usage in any session to see how much of each you have left, or run the free ccusage tool to read your history. The table below shows what each plan costs and how much use it gives, as of November 2026.

Claude Code cost by plan (as of November 2026)

PlanPriceClaude Code?Usage you get
Free$0NoLimited usage, and no Claude Code
Pro$20/month, or $17/month billed yearly ($200 up front)Yes1x: the base amount, every 5 hours and every week
Max$100/monthYes5x Pro usage, every 5 hours and every week
Ultra$200/monthYesNo 5-hour limit for verified accounts; 10x Pro usage every week
Team, Standard seat$25/seat monthly, or $20/seat billed yearlyYesMore usage than Pro
Team, Premium seat$125/seat monthly, or $100/seat billed yearlyYes5x a Standard seat
Enterprise$20/seat/month billed yearly, 20-seat minimum, plus usage at API ratesYesPay for what you use
API key (Console)Pay per tokenYesNo plan windows; spend limits set per workspace

Prices and amounts are from claude.com/pricing in November 2026. Anthropic adds that how fast you reach a limit depends on the model you use, how long your chats are, and how large your files are. So "5x" means five times the Pro allowance, not a fixed number of messages.

If you pay per token with an API key, Anthropic's cost docs give a rough guide: across company deployments the average is about $13 per developer per active day and $150 to $250 per developer per month, with 90% of users under $30 per active day (as of November 2026). Your own number depends on model choice and habits. More on choosing a plan in our Claude Code pricing guide.

How to check Claude Code usage inside a session

The one command to remember is:

/usage

What you see depends on how you pay. Pro, Max and Ultra subscribers see bars showing how much of the 5-hour and weekly windows are used, plus a breakdown. API users see token counts and a dollar estimate for the current session. On the /usage screen:

  • Press d or w to switch the breakdown between the last 24 hours and the last 7 days.
  • The breakdown shows what share of your use came from skills, subagents, plugins and each MCP server (add-ons that connect Claude to other tools).
  • Recommendation rows suggest fixes, such as trimming a long CLAUDE.md file. On recent versions, press f to let Claude Code apply the ones it can.
  • If it says rate limited until a time, only the usage lookup is paused, not your work. Press r to retry after that time.

Two limits on these numbers: they come from session history on this computer only, so use from claude.ai or another machine isn't counted, and the session dollar figure is worked out at list price, so subscribers aren't billed that amount.

Claude Code token usage, live

While Claude works, the line next to the spinner shows a rough count of the tokens (small chunks of text) this turn has written so far, like ↓ 5.5k tokens. To see what fills the context window (how much text Claude is holding in mind), run:

/context

Keep usage on screen with the status line

The status line is a bar at the bottom of Claude Code that shows whatever you like. Describe it in plain words and Claude Code writes the script for you:

/statusline show my 5-hour and weekly usage percentages and context percentage

For subscribers, the status line receives fields called rate_limits.five_hour.used_percentage and rate_limits.seven_day.used_percentage, plus resets_at times for each window, so it can show the same numbers as /usage all the time. They appear after the first reply in a session.

Checking /usage only helps if you remember to. VibeCodingDeck works as a Claude Code usage tracker that never leaves the screen. It keeps a live plan-usage meter on screen at all times, with daily and weekly windows for every account you use, including several logins per AI. Each pane also has a context meter so you can clear a chat before it fills up. When one account runs low, you can hand the job to a different AI and keep going. See what else it does.

ccusage: free usage reports from your own logs

ccusage is a free open-source tool by ryoppippi, under the MIT licence. It reads the log files that Claude Code and other coding agents save on your computer, and turns them into tables of tokens and estimated cost. You don't have to install it. With Node.js on your computer, run:

npx ccusage@latest            # usage by day
npx ccusage@latest weekly     # by week
npx ccusage@latest monthly    # by month
npx ccusage@latest session    # by conversation
npx ccusage@latest blocks     # by Claude Code 5-hour window
npx ccusage@latest codex daily  # Codex usage instead

If you use Bun or pnpm, bunx ccusage and pnpm dlx ccusage work too. Useful options listed in its README:

  • --since and --until to look at a date range.
  • --breakdown to split cost by model.
  • --json for output another program can read.
  • --last 1 to show only the current day, week or month.
  • A statusline command (marked beta) for a compact line in Claude Code's status bar.

Besides Claude Code and Codex, its README says it reads logs from OpenCode, Amp, Gemini CLI and others. Two things to keep in mind. It only sees logs on the computer you run it on, so use from your laptop and desktop won't add up. And its dollar figures are estimates at list prices, which is useful for spotting heavy days but is not your bill on a subscription.

Why you hit the Claude Code usage limit

Anthropic's cost docs explain where usage goes and why people run into Claude Code usage limits sooner than expected. The causes that surprise people most:

  • Long sessions. Claude Code sends your whole conversation with every request. A one-line question in a chat that has been open all day still pays for the whole day.
  • Coming back after a break. The first message after a long pause misses the cache (a store that makes repeated text cheap), so your full history is processed again at full rate.
  • Heavy models and high effort. Opus uses more than Sonnet, and thinking tokens count as output.
  • Subagents and agent teams. Each one runs its own requests. Anthropic says agent teams use roughly 7 times the tokens of a normal session when teammates run in plan mode.
  • Several sessions at once. Every terminal, desktop session and cloud session draws on the same plan windows.
  • Things running while you're away. Scheduled tasks, /loop, messages from your other sessions, and session recaps can all use tokens while you aren't typing.
  • Big setup files. A long CLAUDE.md is sent with every request.
  • Compacting a huge chat. /compact reads the whole conversation to summarise it, which is itself a big request. /clear costs nothing.

If you run several agents, the windows drain faster in proportion. Read running multiple Claude Code agents before you launch a big batch.

Claude Code rate limit messages and what to do

Message you seeWhat it meansWhat you can do
You've hit your session limitYour 5-hour window is used up. It covers all models.Wait for the reset time shown, turn on usage credits, or switch account or AI
You've hit your weekly limitYour 7-day window is used up. Switching models won't help.Same options; the reset may be days away
You've hit your Opus limit (or Sonnet limit)A limit for one model family onlySwitch to a model from another family with /model
A spend limit message (individual, monthly or team budget)Your paid usage credits reached a spend capRaise the cap in Settings, Usage on claude.ai (or ask your admin), or wait for the reset
rate limited until 3:14am (inside /usage)Only the usage lookup is pausedNothing; keep working and press r later

Usage credits let you keep working past a plan limit by paying for the extra. Run /usage-credits to open the settings page on claude.ai, where you turn them on and set a monthly spend cap. On Claude Code v2.1.234 or later, Claude Code can also wait for the reset and carry on with the interrupted task by itself. /rate-limit-options shows the choices when you hit a limit.

Ways to use less Claude Code usage

  • Run /clear when you switch to an unrelated task. Use /rename first so you can /resume the old chat later.
  • Use Sonnet for everyday coding and save Opus for hard planning. Switch with /model.
  • Lower the effort for simple jobs with /effort.
  • Keep CLAUDE.md under about 200 lines, as Anthropic suggests. Move long, task-specific instructions into skills, which load only when used.
  • Turn off MCP servers you aren't using with /mcp. Command-line tools like gh cost less context than MCP servers.
  • Write specific prompts. "Add input checks to the login function in auth.ts" reads far fewer files than "improve this code".
  • Start bigger jobs in plan mode so a wrong approach is caught before it costs a full build. See plan mode.
  • Press Escape the moment Claude heads the wrong way, and use /rewind to go back.
  • Send noisy jobs, like long test runs, to a subagent so only the summary enters your main chat.
  • Keep agent teams small and shut teammates down when they finish.
  • Check /usage once a day and act on the top recommendation row. Each row only appears when it covers at least 10% of your recent use.

More habits that save tokens are in our Claude Code tips.

Usage and limit questions

How do I check my Claude Code usage?

+

Type /usage in a Claude Code session. Subscribers see bars for the 5-hour and weekly windows plus a breakdown of what used them. For history across days, run npx ccusage@latest.

How much does Claude Code cost per month?

+

As of November 2026, Pro is $20/month ($17 billed yearly), Max is $100/month and Ultra is $200/month. On an API key you pay per token; Anthropic reports company averages of $150 to $250 per developer per month.

When does the Claude Code weekly limit reset?

+

The weekly limit is a 7-day window. The limit message shows the exact reset time, and /usage and the status line's resets_at field show it too.

Does switching models get around the usage limit?

+

Not for the session or weekly limit, which cover all models. It does help with a model-specific message such as an Opus limit: switch to a model from another family with /model.

Is ccusage official, and is it accurate?

+

No, it is a free MIT-licensed community tool, not made by Anthropic. Its token counts come from your local logs, but its dollar figures are estimates at list price and only cover the computer you run it on.

Why is my Claude Code usage so high?

+

Usually long chats that were never cleared, Opus left on for everything, several sessions or subagents at once, or a large CLAUDE.md sent with every request. The recommendation rows in /usage point at the biggest cause.

Can I see usage for more than one Claude account?

+

Claude Code shows the account you are signed in with. VibeCodingDeck shows a live daily and weekly meter for every account you add, with several logins per AI.

Sources

Run up to 12 AI coding agents in one window.

Claude Code, Codex, Gemini, Grok and more, side by side. Windows and Mac. From $29 a month.

Get started

Keep reading