"API Error: Server is temporarily limiting requests (not your usage limit)"
Updated
"API Error: Server is temporarily limiting requests (not your usage limit)" means the API applied a short-lived throttle that is unrelated to your plan quota; the message says so in brackets. There is no window to wait for: retry shortly and check status.claude.com if it persists. Since v2.1.199 Claude Code retries it automatically with backoff before showing it. Your session and weekly limits are untouched, and it is not an account-wide reset event.
- Last reset
- 04 SEP 20:08Z
- Status
- OVERDUE
The gap since the last reset is past 1.5x the recent median.
- Next est.
- ANY TIME
- Days since
- 13.1
days
What does "API Error: Server is temporarily limiting requests (not your usage limit)" mean?
The Error reference groups this message under usage limits but explains that it works differently: the API applied a short-lived throttle unrelated to your plan quota. Claude Code tells it apart from a real limit by the absence of the unified quota headers that a genuine limit response carries.
As of v2.1.199 Claude Code retries it automatically with backoff before showing it, whichever way you authenticate. On earlier versions a session signed in with a claude.ai subscription failed the turn on the first occurrence; only API-key and Enterprise sign-ins retried it.
Issue #68502 (2026-06-15) documents an HTTP 529 overloaded_error being rendered with this text and a "Rate limited" suffix, which made parallel sessions and subagents fail and led users to think they had hit a quota. The Claude Platform error reference lists 529 as an overloaded_error, separate from 429 rate_limit_error.
Which window is full?
| Window | Resets | Where to check |
|---|---|---|
| None: a server-side throttle, not a usage window | No reset time; clears as soon as the server stops throttling | status.claude.com for incidents; /usage to confirm your own windows still have room |
What to do when you see it
- Wait briefly and try again
That is the docs' first instruction. Since v2.1.199 Claude Code has already retried with backoff before you see the message.
- Check status.claude.com
The Error reference points there if the message persists; an incident on the status page explains a run of these errors.
- Do not switch models or wait for a reset
The throttle is not tied to your session or weekly window, so neither /model nor waiting for your reset time changes anything.
- Confirm your windows with /usage
If you want to be sure it is not a quota, /usage shows the session and weekly bars; a real limit would show "You've hit your session limit" instead.
When does it reset?
Your own window: nothing here is about your window. The session and weekly limits keep their normal schedule (every five hours, and at a fixed weekly time), and this error does not consume or extend them. If you want to know where you stand, /usage shows both bars.
Account-wide resets: a throttle is also not an official reset. The board above records resets Anthropic announces, each linked to its original post. A run of these errors is more likely an incident on status.claude.com than anything the board would show.
Related errors
- "You've hit your session limit · resets {time}"Claude Code
- "API Error: Usage credits required for 1M context · run /usage-credits to turn them on (they take effect after you restart Claude Code), or /model to switch to standard context"Claude Code
- "Please run /login · API Error: 401 Invalid authentication credentials"Claude Code
FAQ
Is this my usage limit?
No. The message itself says "not your usage limit", and the Error reference explains that the API applied a short-lived throttle unrelated to your plan quota. Claude Code distinguishes it from a real limit by the missing unified quota headers.
Do I need to wait for a reset?
No. There is no window behind this error. Wait briefly and retry; since v2.1.199 Claude Code retries with backoff on its own before showing the message, whichever way you authenticate.
Why did it say "Rate limited" and break my subagents?
Issue #68502 reports an HTTP 529 overloaded_error rendered with this text plus a "Rate limited" suffix, hard-failing parallel sessions and subagents on that version. The Platform error reference lists 529 as overloaded_error, distinct from the 429 rate_limit_error a real quota returns.
How can I tell this from a real limit?
A real limit shows a message that names the window, such as "You've hit your session limit · resets 3:45pm", and /usage shows a bar at 100%. This message has no window and no reset time, and status.claude.com may show an incident.
Sources
- Claude Code docs: Error reference — "Server is temporarily limiting requests"verified 2026-09-18
- anthropics/claude-code issue #68502 (2026-06-15): [Bug] HTTP 529 overloaded_error is rendered as "Rate limited" and hard-fails parallel sessions/subagentsverified 2026-09-17
- Anthropic status page (status.claude.com)verified 2026-09-18
- Claude Platform docs: API errors (401 authentication_error, 429 rate_limit_error)verified 2026-09-18
Want the next Claude reset on your phone?