“You’ve hit your org’s monthly spend limit” in Claude Code, and the two very different reasons you see it

Literally, new requests are being refused because a monthly spend cap on usage credits has been reached. Usage credits are the pay-as-you-go pool that meters at standard API rates once your included plan usage is gone, and a spend limit is a ceiling someone set on how much that pool can bill per month. When the running spend hits the ceiling, Claude, Cowork, and Claude Code stop accepting new requests until the limit is raised or the billing period rolls over. The message points you at the usage settings page because that is where the ceiling lives. The complication, covered below, is that on a solo account with no org this same wording sometimes appears when no spend cap was ever set.

On Team and Enterprise plans, only an Owner or Primary Owner can change it. They go to Organization settings, then Usage (the page the message links to at claude.ai/admin-settings/usage), and either increase the monthly spend limit or approve individual increase requests. On Pro and Max plans you raise or remove your own limit on usage credits with the /usage-credits command directly in the CLI, and Claude Code prompts you to raise or remove it without leaving the terminal, but changing it requires billing access on the account. If you are a member of someone else's org, you cannot lift it yourself no matter what you do locally.

That is the tell that you are not actually at a spend cap. It is a documented mismatch (see Claude Code issue #56035, reported on version 2.1.126 and closed as a duplicate) where the CLI surfaces "you've hit your org's monthly usage limit" while the web dashboard shows most of the period unused. The logs in those reports show 429 rate-limit responses and stale local lock or cache state, not an exhausted budget. The fix is not to visit an admin page; it is to wait a few minutes for the backend to resync, restart Claude Code, or upgrade to a newer version.

Because the message is partly a naming bug. Claude Code issue #52908 reported that starting in version 2.1.119, sending a message after your session usage was exhausted produced "you've hit your org's monthly usage limit" even for Pro users who have no organization. Issue #52679 and #52960 describe the same confusion. The included usage on Pro and the Max tiers refreshes on a roughly five-hour cycle, so the real meaning in that case is "this five-hour window is used up," not anything about an org or a spend cap. Waiting for the window to reset clears it.

They produce different banners and need different fixes. "Out of extra usage" means the prepaid overage pool emptied and the admin has to top it up or re-enable it. A spend limit is a deliberately set ceiling that stops spending before the pool is empty, so the credits may still be there, the cap is just refusing to spend past the number someone chose. We cover the out-of-extra-usage case in a separate write-up; if your banner is about extra usage rather than a spend limit, start there.

Yes, if you have a separate funded path. Two work: sign in to a tool with your own personal Claude Pro or Max account so usage hits your plan instead of the capped org, or route requests through an Anthropic-API-compatible endpoint you control (a corporate proxy, a gateway, or a local bridge) that does not sit behind the org's Console spend cap. The web chat will not let you swap accounts mid-session, but desktop tools that wrap the Claude Code agent loop generally do. Fazm exposes both: an account switch and a custom endpoint field that overrides the API base URL.

Raising the limit does not change the rate; it just allows more spend to occur. Everything past your included plan usage meters at standard API rates whether the limit is 50 dollars or 5,000. Enterprise deployments average around 13 dollars per developer per active day on Claude Code, and a single long agent run that streams a large repo and many tool results is what tends to push a team over a conservative monthly cap before anyone notices the slope.

Not if your tool keeps sessions on disk. The upstream CLI stores sessions under ~/.claude/projects, and a desktop wrapper that auto-restores windows lets you resume the same conversation under whichever account is still funded. The unit of work is the session and the file changes are in git, so switching the account that pays for inference does not have to mean starting the task over.