Claude Code recognises when your organization's usage is paused by a daily cap

What

Claude Code can now tell when usage is paused for your whole organization because of a daily cap. It treats this differently from an ordinary rate limit, which is a limit on how much you can use in a period.

• The rate-limit reader checks a new response header, anthropic-ratelimit-unified-overage-daily-cap-paused. When it is true, it sets dailyCapPaused on the limit status, including on refused requests.
• When an API rate_limit error carries quotaLimits.dailyCapPaused set to true, the pause gets the cause daily_cap instead of usage_limit or rate_limited. Which parts of Claude Code act on the new cause was not traced.
• A paused /goal now reads "Goal paused · usage is paused for your organization". A message "Usage is paused for your organization. For help, please contact support." also appears.
• The usual rate-limit message is skipped when dailyCapPaused is true.
• While the cap is paused, the offer to buy extra usage does not appear and the automatic arming for rate limits stops. The extra-usage flow also stops early when daily_cap_notice is set.
• A plan or limit check now also requires dailyCapPaused not to be true.
• In the api_error description, rate_limit_info for usage_limit_reached now includes dailyCapPaused, which is true when usage is paused for the organization. Programs built on the SDK (the kit for driving Claude Code from your own code) can read it.
• A quota error is now ignored when the caller passes onlyLimiterRefusals and the status is not rejected.
• The client defines a new header name, and how it is used was not traced.

Separately, the spend-limit proxy now tries a second read when the caps come back empty. Its 429 (too many requests) message now says "spend limit reached" with a reset date.

Why

If your organization hits a daily cap, you get a clear reason, and you are not offered extra usage that would not help. Programs on the SDK can tell this pause apart from other limits.