On a Pro or Max plan, Claude Code has two limits: a session limit that resets every five hours, and a weekly limit. Both are shared with the Claude apps, so chatting in Claude uses the same allowance as coding in Claude Code.
The limits, by plan
| Plan | 5-hour session limit | Weekly limit | Separate weekly limit for Fable |
|---|---|---|---|
| Pro | Yes | Yes | No |
| Max | Yes, 5 times Pro | Yes, 5 times Pro | Yes |
| Ultra | No (verified accounts) | Yes, 10 times Pro | Yes |
Prices for each plan are in how much Claude Code costs.
Anthropic does not publish these limits as a fixed number of messages. Its help pages say the amount you get varies with your plan and with what each message costs, so the honest way to know where you stand is to look at your own percentage.
Everything counts against the same limits
The session and weekly limits are shared across Claude and Claude Code. Chat in the web, desktop or phone app, Claude Code in a terminal, and Claude Code in an IDE all draw from one allowance. A heavy afternoon in the chat app leaves less for a coding session that evening.
When they reset
- Session limit: resets every five hours.
- Weekly limits: reset once a week.
The exact time is per account, and Anthropic's pages do not give a rule for it. You do not need one: the reset
time for each limit is shown next to its bar in Settings > Usage and in /usage.
Buying usage credits does not move the reset time.
What uses the limits fastest
Anthropic lists these as the things that make a message cost more:
- How long the conversation already is. The whole conversation is sent again with every request, so a long session costs more on every turn, not once.
- The model, and the effort level. Anthropic's own advice is to use Sonnet for most work and keep the larger model for the hard parts.
- Tools: web search, research, running code, creating files.
- Large attachments and long messages.
- Subagents and agent teammates, because each one makes its own requests.
- Scheduled tasks and
/loop, which keep spending while you are away.
One thing works in your favour: content that is reused soon afterwards is cached and counts less. A session left idle loses that cache, so the first reply after a long break is a costly one.
In practice, three patterns did most of the damage on the author's Mac: sessions that start large because of connectors and plugins, sessions left to grow instead of being compacted, and heavy use of subagents.
How to check where you stand
/usagein Claude Code: bars for your plan limits, plus a breakdown of what counted against them over the last 24 hours or 7 days./statusin Claude Code: your remaining allocation.- Settings > Usage at claude.ai: the current session and the weekly limits, with the time left.
- A status line: Claude Code can show both percentages at the bottom of the terminal as you work. How to set one up.
More detail on each is in how to check your Claude Code usage.
What happens when you hit a limit
If a reply is in progress when you reach the limit, Claude Code may finish it; you will see "Usage limit reached ยท wrapping up". Pro gets this once per weekly period, Max each time a session limit is hit. After that you can:
- Wait for the reset. Recent versions of Claude Code can wait and continue the task by themselves once the limit resets.
- Turn on usage credits. You keep working and the extra is billed at standard API rates, up to a monthly cap you set.
- Upgrade from Pro to Max, or Max to Ultra.
- Use an API key from a Claude Console account and pay per token.
Switching to a different model helps only when the limit you hit is the one for a single model. The session limit and the all-models weekly limit apply whichever model you pick.
Know before you hit it
The percentage alone does not tell you whether you will make it to the reset. 60% used with four hours left is a problem; 60% used with twenty minutes left is not. What you want is the pace: at this rate, when do you run out, and which session is responsible?
Cost Doctor shows both limits in your menu bar with that forecast ("at this pace you run out at 3:35 PM, 35 minutes before it resets") and names the session that used most of the current window. The free costdoctor command prints the same forecast in a terminal.
Plan facts on this page are from Anthropic's help pages and Claude Code documentation, as checked on November 10, 2026: Use Claude Code with your Pro, Max, or Ultra plan, Usage limit best practices, Manage usage credits and Manage costs effectively, and the plans page. Limits change; check those pages for the current rules.
See it on your own sessions
Cost Doctor is a Mac menu bar app that reads the logs Claude Code and Codex already write. It names the session that is using your tokens, says why, and turns your screen red when a check fails while an agent sits waiting. Nothing leaves your Mac.
Download the free trialHow it works14 days free, no account. Then $19 once, plus tax where it applies. For Apple silicon and Intel Macs on macOS 13 or later.