DEV Community

Cover image for Claude Code hit its usage limit? How the limits work, and how to keep working
Roi Dayan
Roi Dayan

Posted on Fully Autonomous

Claude Code hit its usage limit? How the limits work, and how to keep working

#ai #claudecode #codex #productivity

"You've hit your session limit" is the line nobody wants mid-task. Here's how Claude Code's limits actually work on a Claude plan, how to see one coming, what your options are when you hit it, and how to make your usage last longer. Every fact is from Anthropic's own pages, checked on November 8, 2026.

How the limits work

  • Two windows on Pro and Max. "Pro and Max plans have a five-hour session limit and a weekly limit. Ultra plans have a weekly limit, and verified Ultra accounts don't have a five-hour session limit." (Anthropic)
  • The five-hour window starts with your first message. "The period starts when you send your first message, and your limit resets in full five hours later." (Anthropic)
  • The weekly limit resets at a fixed time assigned to your account, whenever you started using Claude. It applies across all models, and some plans also have a separate weekly limit for Fable.
  • Chat and Claude Code share one pool. "All activity in both tools counts against the same limits." A long chat on claude.ai in the morning is less Claude Code in the afternoon.
  • The plans: Pro, Max at five times Pro's usage, and Ultra at ten times Pro's weekly usage. Ultra is what was called Max 20x. (Max, Ultra)

See it coming

  • /usage in Claude Code shows your plan's usage limits, with session cost and activity stats. (Commands)
  • On the web, Settings > Usage shows your current session and when your weekly limits reset.
  • A warning before the wall: Claude Code can tell you when you've used most of a limit, with a line such as You've used 85% of your session limit · resets 3:45pm. (Errors)
  • Your status line can show both windows: for subscribers, its rate_limits object carries a five-hour and a seven-day window. (Status line)

When you hit it

At the limit, Claude Code says which one, such as You've hit your session limit · resets 3:45pm, and blocks further requests until that time. Your options:

  1. Wait, and let it continue on its own. With a claude.ai subscription, automatic continue is on by default: Claude Code shows Continuing automatically at 3:45pm · Esc to cancel and picks up at the reset. Turn it off in /config with Continue automatically at usage limit. (Interactive mode)
  2. See every option in one place: /rate-limit-options lists waiting and continuing at the reset, adding usage credits, and upgrading your plan.
  3. Turn on usage credits in Settings > Usage, billed at standard API rates, with a monthly spend limit you set. In Claude Code, /usage-credits does it (it used to be /extra-usage). (Usage credits)
  4. Use an API key from the Claude Console for a heavy sprint, billed at API rates. Mind the reverse too: if ANTHROPIC_API_KEY is set in your environment, Claude Code uses it instead of your subscription. Check with /status. (Authentication)
  5. Switching models won't help with the session or weekly limit, which are shared across all models. It does help with a model's own limit, such as Fable's weekly one.

Make it last longer

From Anthropic's own cost guide:

  • /clear between unrelated tasks. "Stale context wastes tokens on every subsequent message." It costs nothing.
  • /compact with a focus, such as /compact Focus on the API changes. It reads the whole conversation to summarize it, so it's a big request itself: use /clear when you don't need the history.
  • Pick the model for the job. The default on Pro, Max and Ultra is Opus. "Sonnet handles most coding tasks well and costs less than Opus." /model opusplan plans with Opus and builds with Sonnet.
  • Keep CLAUDE.md under 200 lines. It's read into every session; move workflow-specific instructions into skills.
  • Subagents count too. Their requests draw on the same limits. Give simple ones model: haiku.
  • Lower the effort with /effort for routine work, and turn off MCP servers you aren't using (/mcp). /context shows what's filling the context.

Keep the work going with another agent

The limits are per plan, so a second agent on a different plan keeps the work moving. Codex on a ChatGPT plan has its own limits, with a five-hour window on Plus, and Pro has none for now. Its pricing page says /status shows what's left during a session, and there's a usage page at chatgpt.com/codex/settings/usage. (Codex pricing)

Handing work across is easy if the work is already set up for it: a short AGENTS.md that both agents read, one task per branch, and a handback that names the commit, so the next agent starts from exactly where the last one stopped.

Doing this in Moontors

I build Moontors, a free desktop app that runs Claude Code, Codex, Gemini CLI and Grok Build side by side, so take this part as the maker's view.

  • Your limits where you can see them. With a Claude subscription, a ChatGPT sign-in for Codex or a paid grok.com plan for Grok Build, Moontors shows each limit's window, how much is used and when it resets. From 75% a ring appears in the header, amber, then red from 90%. (Context, usage and limits)
  • Claude Code and Codex on your own plans, Claude Code on your Claude subscription and Codex on your ChatGPT sign-in.
  • Agents that hand work to each other. A Claude Code session can hand a task to a Codex session and hear back, so a limit on one plan doesn't stop the work.

AI did the writing, I did the reading, and we both took credit.

Top comments (0)