morphy
Download

Claude Code usage limits: how to stop hitting them (2026)

How Claude Code's 5-hour and weekly limits work on Pro and Max, what to do when you hit one, and 12 ways to make your plan last longer, from Anthropic's own docs.

Casper Apers6 min readFacts checked October 16, 2026

You're in the middle of something and Claude Code stops: You've hit your session limit · resets 3:45pm. This guide explains how the limits work, what to do when you hit one, and the habits that make a plan last longer. It's all from Anthropic's own documentation, checked on October 16, 2026.

(I build Morphy, which runs Claude Code next to other AIs. It gets one short section near the end; everything else applies whichever app you use.)

How do Claude Code's usage limits work?

  • Two limits on Pro and Max: a 5-hour session limit and a weekly limit. Both count at the same time, so one heavy burst, like a big fan-out of subagents, can use up the week before the session window resets.
  • Shared with Claude chat. Everything you do on claude.ai counts against the same limits as Claude Code.
  • Two model limits on top: an Opus limit and a Sonnet limit, which only apply to that model family.
  • Max gives five or twenty times Pro's allowance per session ($100 or $200 a month). Its monthly API credits can't be used in Claude Code or the Claude apps, only on the API.
  • Team and Enterprise give each person a seat allowance on the same 5-hour and weekly windows, shared with Claude chat and Cowork; a premium Team seat has five times a standard one.
  • The free plan doesn't include Claude Code at all.

When you run out, Claude Code tells you which limit it was:

MessageWhat it meansWhat helps
You've hit your session limit · resets 3:45pmThe 5-hour allowance is used up, for every modelWait, usage credits, or another AI
You've hit your weekly limit · resets Mon 12:00amThe week's allowance is used up, for every modelThe same; switching models won't help
You've hit your Opus limit (or Sonnet limit)Only that model family is used upSwitch to another family with /model
You've used 85% of your session limitA warning before the limitFinish the step you're on, plan the next

If Claude Code is in the middle of a step when the limit hits, it may keep working briefly to finish it.

How do you see how much you have left?

  • /usage shows your limits and when they reset. On a plan, it also shows which skills, subagents, plugins and MCP servers used the most, and flags habits that cost you, like a long context or cache misses. Press d or w for the last day or week.
  • The status line can show your remaining allowance all the time (its rate_limits fields).
  • The Claude desktop app has a usage ring next to the model picker, and the VS Code extension an Account & usage dialog.

What can you do when you hit the limit?

  1. Let it carry on by itself. Since version 2.1.234, Claude Code waits in the open session and continues the task on its own after the reset; it's on by default when you're signed in with a Claude plan. In the desktop app, tick Auto-continue when limits reset on the session-limit card.
  2. Switch models, if it was the Opus or Sonnet limit. The next message re-reads the whole conversation without the cache, so it costs a little more.
  3. Turn on usage credits. /usage-credits opens Settings → Usage on claude.ai, where you can keep working past the limit at standard API prices, with a monthly cap you set. Note that while you're on usage credits, Claude keeps your conversation in its cache for five minutes instead of an hour, so pauses cost more.
  4. Upgrade to Max, for five or twenty times Pro's allowance.
  5. Hand the task to another AI. If you also have ChatGPT, Codex reads the same AGENTS.md and runs on a separate plan with its own limits. See Codex vs Claude Code.

12 ways to make your plan last longer

In a long session, much of the usage is Claude re-reading the conversation: every request sends it again, and every tool call adds to it. So the biggest savings come from keeping the context small.

  1. Clear between tasks. /clear starts fresh and costs nothing. Use /rename first if you want to come back to the old session with /resume.
  2. Use Sonnet for most work. It handles most coding well and uses less than Opus; keep Opus for hard design decisions. Switch with /model, and remember that subagents that inherit your model switch with it.
  3. Lower the effort for simple jobs. The 5.5 models always think, but /effort sets how much.
  4. Mind the cache. On a plan, Claude keeps your conversation cached for an hour. After a longer break, your next message re-reads all of it at the full rate, so a long break is a good moment for /clear.
  5. Compact on purpose. /compact reads the whole conversation to summarize it, so it's a big request itself. Tell it what matters (/compact Focus on the failing tests), or /clear when you don't need the history.
  6. Keep CLAUDE.md short. It's loaded into every session. Move instructions for specific jobs, like releases or migrations, into skills, which load only when needed.
  7. Trim your MCP servers. /context shows what's filling the context, and /mcp turns off servers you're not using. Command-line tools like gh cost less context than an MCP server that does the same.
  8. Send noisy work to subagents. Test runs and long logs stay in the subagent's context, and only a summary comes back. Subagents still use your plan, though, so give them a smaller model.
  9. Keep agent teams small. Teammates in plan mode use about seven times as much as a normal session, because each one is a separate Claude with its own context.
  10. Plan before big changes. Plan mode (Shift+Tab) agrees on an approach before any code is written. Press Esc as soon as it heads the wrong way, and /rewind to go back.
  11. Be specific. "Add input validation to the login function in auth.ts" reads a few files. "Improve this codebase" reads everything.
  12. Watch what runs in the background. A scheduled task or loop sends your whole conversation each time it fires, even when you're not there. A hook can trim noisy output, like keeping only the failures from a test run, before Claude reads it.

What about running several agents at once?

Every session and subagent draws from the same plan, so five agents at once use it about five times as fast. Two things help: give each agent a small, separate task with its own clear context, and put some of them on a different plan or a free model, so one limit doesn't stop everything.

In Morphy

Morphy runs Claude Code with your own sign-in, next to Codex, Copilot, Gemini, Grok and others. Settings → Plan usage shows how much of your Claude, Codex and Antigravity plans you've used, and when each resets. When a chat hits its limit, it offers Send when it resets (on Pro) and Continue with Codex, or any other AI you've connected: a new chat beside it, in the same folder, that carries the conversation on. Small jobs can run on OpenCode's or OpenRouter's free models, or Ollama on your own computer. See Connect your AI.

Questions

When does the limit reset? The message tells you ("resets 3:45pm"), and /usage shows both your session and weekly reset times.

Does using Claude chat count against Claude Code? Yes. On Pro and Max, both draw from the same limits.

Can I use Max's API credits in Claude Code? No. They're for building on the Claude API, not for Claude Code or the Claude apps.

Does switching to Sonnet help? Only for an Opus limit. The session and weekly limits cover every model.

Is it cheaper to use an API key? It depends on how much you use. Usage credits and API keys are both billed at standard API prices, while a plan costs the same every month. If you hit your limits often, compare a month of API spend with Max; for occasional use, paying per use can be cheaper.

References