Is my Claude Code usage limit a bug?

Rule out the usual causes first

1. Read the message. It names the limit you hit: session, weekly, Opus, Sonnet or your monthly spend limit, and for a session, weekly or model limit, when it resets. Every request counts against the session and the weekly limit at once, so the week can run out first.
2. Count everything on your account. claude.ai, the desktop app, your IDE and every other computer share the same limit. The breakdown in /usage only counts sessions on this computer, so check Settings > Usage on claude.ai for the whole picture.
3. Check which account you're on. /status shows the account and whether an API key is in use. An ANTHROPIC_API_KEY in your environment bills the API instead of your plan. Unset it, remove it from your shell profile and start Claude Code again, then run /login if you aren't signed in with your plan.
4. Check the model and effort. The default on paid plans is Opus 5.5, which uses your limit faster than Sonnet (its API price per token is twice Sonnet's), and higher effort means more thinking billed as output.
5. Look for work you didn't watch. Subagents, agent teams, scheduled /loop tasks and idle compaction all spend while you're looking elsewhere.
6. Make sure it's a limit at all. A 529 overloaded error doesn't count against your quota, and "Server is temporarily limiting requests (not your usage limit)" isn't your plan's limit either. Check status.claude.com for incidents.

Anthropic also says it may adjust limits, so work that used to fit in a week may not now. (Anthropic: how usage limits work, Claude Code errors)

Known usage limit bugs

These are live on the Claude Code RAP Sheet, grouped from public GitHub issues in anthropics/claude-code. Each page shows how many people hit it, whether it's growing and any known workaround.

• Usage limits drain too fast: the weekly limit runs out in a day or two on work that used to last the week. It includes:
  • Higher tiers deplete limits faster
  • Usage meter full after quota reset
  • Weekly limit wrong after upgrade
  • Fable meter rises without Fable use
• Credits ignored at usage limit: usage credits are on, but Claude Code still says you've hit your limit, or they're billed while the weekly limit still blocks you.
• Unrequested agent fan-outs drain usage: Claude starts many subagents for a small task and burns through the limit.
• Auto-continue doesn't resume after the reset: it's on in /config, but the session never carries on when the limit resets.
• Session limit kills running agents: hitting the session limit stops running workflow agents instead of pausing them. Its page has an unverified workaround.
• Usage stuck on old account: after switching accounts, the limits stay tied to the old one.

If yours isn't listed

Search the Claude Code RAP Sheet for your exact error; it takes in new reports about every 15 minutes. If it's not there, report it to Anthropic with /bug in Claude Code, which can include your session history once you confirm.

More guides

• How do Claude Code usage limits work?
• How do I reduce Claude Code token usage?

Facts about limits and commands come from Anthropic's own pages, checked Nov 2, 2026; when this page and theirs differ, theirs is right. The bugs above are written in by hand; the RAP Sheet has the current list.