Claude Code
Claude Code usage vs API usage
Claude Code can run on two different billing models. Signed in with a Pro or Max subscription, it draws on a plan allowance: a five-hour session limit and weekly limits, shared with claude.ai and Claude Desktop. Signed in with an API key, you pay per token, there are no plan windows, and the API's own rate limits apply instead.
| Subscription (Pro, Max) | API key | |
|---|---|---|
| How you pay | Flat plan price | Per token used |
| What limits you | Five-hour session limit and weekly limits | Spend you allow and API rate limits |
| Shared with | claude.ai and Claude Desktop | Only other uses of that key |
| Where to look | /usage, status line, Usage Meter | The Anthropic Console |
| When it runs out | Stops until the window resets | Rate limits clear in moments; spend limits are yours to raise |
How to tell which you are on
If Claude Code reports a five-hour and weekly percentage, you are on a subscription: those limits are shown only to Pro and Max subscribers. If it reports no plan windows and you signed in with an API key, you are billed per token. Anthropic's cost documentation covers the details of each.
Why it matters for tools
- The Usage Meter and Paneherd's limit display show subscription windows. On an API key they have nothing to show.
- Parallel sessions spend either one faster: a plan window or your budget. See how parallel sessions use the limits.
- The same split exists in Codex: ChatGPT sign-in has windows, an API key is per-token. See Codex CLI usage limit.
Frequently asked questions
Can I switch between them?
Yes, by signing in differently. Limits and billing follow whichever you are signed in with.
Which hits a limit first?
They are not comparable: a plan stops at its window, an API key stops at your budget or at a rate limit. See rate limit vs usage limit.