Guide

Claude Code usage limits, explained

The 5-hour window, the weekly cap, how to check both, and what to do when Claude Code stops.

Last updated: November 15, 2026

Claude Code on a Pro or Max plan has two usage limits. A 5-hour session limit starts with your first message, and a weekly limit resets at a fixed time each week. Both are shared with claude.ai and Claude Desktop.

Anthropic does not publish token numbers for either window, and it can change them. So this guide sticks to how the windows behave, how to read your own numbers, and what to do when one runs out. Every fact links to Anthropic's docs as read on November 15, 2026.

What are Claude Code's usage limits?

Every subscription plan has a weekly limit. Pro and Max also have a 5-hour session limit, and Max and Ultra add a separate weekly limit for Fable models. Usage in Claude Code, claude.ai, Claude Desktop and your IDE all counts against the same limits (Anthropic).

Plan5-hour limitWeekly limitNotes
ProYesYes, across all modelsWrap-up of the current step up to once per week
MaxYesYes, plus a separate Fable weekly limitWrap-up each time you reach the 5-hour limit
UltraNone for verified accountsYes, plus a separate Fable weekly limitA temporary limit may apply to unverified accounts
Team and Enterprise seatsRolling 5-hour windowWeekly windowShared with Claude chat and Cowork; size depends on seat tier

Sources: plan limits, Team and Enterprise seats. On May 6, 2026 Anthropic doubled Claude Code's 5-hour limits for Pro, Max, Team and seat-based Enterprise plans (announcement).

How does the Claude Code 5-hour limit work?

The 5-hour window opens when you send your first message and resets in full five hours later. Work that starts at 10:00 frees up again at 15:00, however much of the window you used (Anthropic).

The window is per account, not per terminal. Three Claude Code sessions and a claude.ai chat all draw from the same 5 hours. Verified Ultra accounts have no 5-hour limit.

How does the Claude Code weekly limit work?

The weekly limit covers all models and resets at a fixed weekly time tied to your account. The reset time does not move with your first message or your subscription date (Anthropic).

Some limits are per model. Claude Code can stop with You've hit your Opus limit or You've hit your Sonnet limit while the weekly limit still has room (error reference). The usage data Clauder reads reports a separate weekly Sonnet counter next to the all-models one, so check both.

What uses up your Claude Code limit faster?

Context size moves usage most. Claude Code sends the whole conversation with every request, and each tool call sends another one, so a one-line question in a session open since morning still pays for the full history (Anthropic). The other drivers:

  • The model and effort level. Opus costs more per request than Sonnet.
  • Your first message after a long break misses the prompt cache and reprocesses everything.
  • Subagents, agent teammates and scheduled loops send requests of their own.
  • /compact reads the conversation it summarizes, so it is a large request. /clear costs nothing.

How do I check my Claude Code usage?

Run /usage. On a subscription it shows your plan usage bars and a breakdown of what counted against them, with suggested fixes. Press d or w for the last 24 hours or 7 days. The breakdown only covers sessions on this machine (Anthropic).

For an always-on number, a custom status line receives rate_limits.five_hour and rate_limits.seven_day, each with used_percentage and resets_at. They appear for Pro, Max and Ultra subscribers after the first response of a session (status line docs):

Bash
jq -r '.rate_limits.five_hour.used_percentage // empty'

On claude.ai, Settings > Usage shows time left in the current session and the next weekly reset.

Watch for one trap. The usage check has its own rate limit. If every terminal polls it, the check starts failing with HTTP 429 and /usage falls back to the last bars it loaded in the past hour. A "rate limited until" message there is about the readout, not your plan (Anthropic). Clauder hit this with several terminals open. It now lets one process fetch for all of them behind a file lock and waits out the server's Retry-After, capped at 5 minutes.

What happens when you hit the Claude Code limit?

Claude Code stops and shows You've hit your session limit or You've hit your weekly limit with the reset time (error reference). If it was mid-response, it may finish the step it was on and show "Usage limit reached · wrapping up". Pro gets that up to once per weekly period, Max each time it reaches the 5-hour limit (Anthropic).

Switching models with /model does not help with a session or weekly limit, because those cover every model. It does help after a model-specific Opus or Sonnet limit (Anthropic).

How do I keep working after hitting the limit?

You have four options (Anthropic):

  • Wait for the reset. On Claude Code v2.1.234 or later it can resume the interrupted task on its own once the window resets. See /rate-limit-options.
  • Turn on usage credits. Run /usage-credits. Requests then bill pay-as-you-go until your monthly spend limit.
  • Upgrade the plan. Pro to Max, or Max to Ultra.
  • Use API credits. Sign in with a Claude Console account and pay standard API rates. The monthly API credits bundled with Max and Ultra don't work in Claude Code.

How do I see the limit coming before it hits?

Claude Code tells you how much you have used, not when you will run out. Clauder adds the missing part. It reads the same usage numbers on your machine, shows both windows in the VS Code or Cursor status bar and the Claude Code statusline, and on Pro predicts the time you will hit the limit from your own usage pattern.

  • Free: both limits in your status bar, cost by model, basic budget alerts.
  • Pro ($1.99/month or $14.99/year): limit predictions plus alerts to Slack, Discord or a webhook before you hit the wall.
  • No Anthropic API key, no prompts uploaded. Only token counts, models and timestamps sync.

Install Clauder free or compare plans.

Where do these facts come from?