Claude Code
Why am I rate limited in Claude Code?
You are usually spending the allowance faster than your own typing suggests. On a subscription, Claude Code, claude.ai and Claude Desktop all draw on one five-hour limit and weekly limits, and every extra session, long conversation or forgotten scheduled task draws on it too. Work through the checklist below; /usage shows which cause is yours.
Checklist
- Which limit is it? A plan usage limit names a reset time. An API rate limit (429) or an overloaded error does not. See rate limit vs usage limit.
- Check the numbers. Run
/usage, or look at the five-hour and weekly percentages in the Usage Meter. Whichever is near 100% is the one that stopped you. - Count your sessions. Every open session draws on the same allowance. Four running at once spend it roughly four times as fast. How parallel sessions use the limits.
- Look for a long conversation. Claude Code sends the whole conversation with each request. A session open all day costs far more per prompt than a fresh one;
/clearwhen you switch tasks. - Look for background spend. Forgotten
/looptasks, subagents and agent teams keep using tokens./usagelists them. - Other apps count too. Time in claude.ai or Claude Desktop is part of the same limit.
- Check the model and effort. Larger models and higher effort use more.
Each of these is covered in why Claude Code usage runs out so fast.
Seeing it before it happens
The free Usage Meter warns you at a percentage you choose, 85% by default. If you run several agents, Paneherd also estimates each pane's share of the current window, so the session that is spending most stands out.
Frequently asked questions
When will I be able to use Claude Code again?
When the window resets. How to find the reset time.
Can I keep working meanwhile?
Another agent bills its own plan, for example Codex. Paneherd runs both in one VS Code grid.