Claude extra usage: how pay-as-you-go overflow works

Claude extra usage, which Anthropic now calls usage credits, lets a paid plan keep working after it hits its 5-hour or weekly limit. You turn it on in Settings › Usage, prepay some funds and set a monthly cap. Past the limit, Claude and Claude Code carry on and you pay standard API rates for that usage, until the cap or your balance runs out.

On a heavy day that is a safety net. Paid every month, it can cost more than the next plan up, and the line between the two is the price gap: $80 a month from Pro to Max 5x, $100 from Max 5x to Max 20x.

How Claude extra usage works#

Anthropic's help page says usage credits "allow individuals subscribed to paid Claude plans (Pro, Max 5x, and Max 20x)" to keep using Claude after reaching their included limits, and they "are billed at standard API rates" (Claude Help Center). The page's address still says "extra usage", which is the name most people search for.

On 22 October 2026 the help page and the Claude Code docs set five rules.

• Claude Code counts: "Usage credits apply to both Claude conversations and Claude Code terminal usage."
• You prepay. Auto-reload, if you turn it on, tops the balance up when it falls below a threshold you set.
• You pick a monthly cap ("Set a maximum amount you're willing to spend on usage credits each month") or "Set to unlimited". /usage in Claude Code then shows this month's spend against the cap, or Unlimited and no spend figure.
• Spending stops at the cap, or at an empty balance with auto-reload off: "Your usage continues until you reach your monthly spend limit or your usage credit balance runs out with auto-reload off."
• Your plan's included usage goes first. A few requests are paid only from credits, such as a model or a 1M context window your plan includes only through them (Claude Code docs).

On Team and Enterprise plans, admins turn usage credits on for the organization and can set spend limits for the whole organization, a group or one member (Claude Code docs).

What it costs#

Usage past the limit is billed per token at the API price of the model you use. As of 22 October 2026, in USD per million tokens (Claude docs):

Credits also change the cache. Claude Code writes the main conversation's prompt cache for one hour on a subscription and for five minutes on usage credits (Claude Code docs). On Opus 5.5 each write then costs $5 per million tokens instead of $8, and after a pause of more than five minutes the next request re-reads the whole conversation. Setting promptCacheTtl to 1h, or CLAUDE_CODE_PROMPT_CACHE_TTL=1h, keeps the hour (Claude Code v2.1.242 or later).

How to turn it on#

1. On claude.ai, open Settings › Usage.
2. In Usage credits, click Enable.
3. Add a payment method if you need to, and prepay some funds.
4. Click Adjust limit and set a monthly spending cap. Don't leave it unlimited unless you mean it.
5. Optional: turn on auto-reload with a threshold.

From Claude Code, /usage-credits does the same: on Pro or Max it opens Settings › Usage in your browser; on Team or Enterprise it opens your organization's usage settings, or sends your admins a request if you have no billing access. It needs a claude.ai sign-in and isn't available with API key authentication. Once credits are on, /usage shows this month's credit spend against your cap (Claude Code docs). If you subscribe through the mobile app, the help page says you can only enable credits on the web.

When it beats a bigger plan, and when it doesn't#

Credits are overflow, not a plan. A release week, a migration or one long Friday that runs past the limit gets billed for those hours and no others, and the cap bounds the bill.

When a credits bill turns up every month, put it next to the next plan's price. As of 22 October 2026, from Claude pricing and Anthropic's Max plan help page:

If your credits regularly cost more than the gap to the next plan ($80 from Pro to Max 5x, $100 from Max 5x to Max 20x), the bigger plan is likely the better deal. $80 at the rates above is 4 million Opus 5.5 output tokens, and the usage-credits row in /usage gives you the month's figure. The weekly limit explains when the week resets, and API key or subscription? compares a key with a plan on your own numbers.

A second account gives you a second set of windows rather than overflow billing. Use only accounts you own and are entitled to use; providers' terms may not allow sharing a subscription sign-in. The other options at a limit are in Claude limit reached: what to do.

How Shuriken handles overflow#

Usage credits are a setting on your Claude account. Shuriken has its own, separate idea of overflow: Switch to API keys, which is off by default.

Turn it on, and API keys and endpoints you've added for that agent's own provider take chats only once every subscription of that agent is out. A spend cap per fallback run, on by default at $25, stops the run. It counts everything the run spends for that agent's provider, across every key it uses, not per key. As soon as a subscription has room again, the run ends, and the next one starts from a fresh cap. We wrote up why it works that way in API keys as a last resort.

One thing sits outside that switch. If you added an endpoint for Claude Code that serves another company's models, Kimi or GLM for example, a chat can go on there between turns once every Claude source is out. That is on by default, and the spend cap above doesn't apply to it.

Before any of that, hot swapping moves your chats between the subscriptions you've added, so overflow is the last stop, not the first. Download Shuriken.