Claude extra usage, explained and under control
Past your plan limits, every prompt bills real money at API rates. Anthropic caps that spending once a month. tokenkarma shows the burn live and blocks the Claude Code prompt before it spends, at your number: per session, per day or per week.
Extra usage lets you keep working past your Claude plan's included limits, billed per token at standard API rates on top of your subscription. Anthropic's only spend control on it is one monthly cap.
Since July 7, 2026 it matters more: Fable 5 is no longer included in any plan and runs entirely on extra usage, at $10 per million input tokens and $50 per million output tokens. tokenkarma shows the burn live and blocks Claude Code spend at the limit you set: per session, per day or per week.
What extra usage is
A Claude plan comes with included usage windows: the 5-hour session and the weekly cycle. Extra usage (Anthropic also calls it usage credits) is the meter that starts when you opt to keep going past them: your prompts continue, and each one is billed per token at standard API rates, on top of what you already pay for the plan.
It is genuinely useful: no more waiting for a reset at 2pm on a deadline. It is also the moment your subscription quietly grows a variable bill, and the meter does not pause to ask whether you meant it.
What it costs
Standard API rates, per model. The one to know since July 7 is Fable 5: $10 per million input tokens, $50 per million output tokens, exactly twice the Opus 4.8 rate. By Anthropic's own docs it also tokenizes roughly 30% more tokens for the same content, so the effective gap versus Opus is wider than the sticker suggests. Cheaper models bill proportionally less: the model mix, not the hours, decides the bill.
- Until July 7, 2026 Fable 5 was included in paid Claude plans, up to 50% of your weekly usage limits.
- Since July 7 Every Fable 5 prompt bills usage credits: $10 per million input tokens, $50 per million output tokens. Twice the Opus 4.8 rate, on top of your plan.
- When credits run out Fable 5 access stops mid-session. No automatic fallback to a cheaper model.
Anthropic frames the change as temporary, "when sufficient capacity allows", with no timeline. Facts as published by Anthropic, July 2026.
See the burn while it happens
The whole problem with extra usage is that it is invisible until the invoice. tokenkarma makes it visible while you work:
- a red extra-usage alert in the Mac menu bar island the moment the meter starts,
- a live burn ticker: the dollars, moving, in your menu bar,
- the dashboard overage card with your monthly figure next to every other AI you pay for,
- an over-budget alert before the number becomes a story.
Seeing all of this is free, on every plan. The blocking below is where Expert starts.
Block it before it spends
Type /tokenkarma 20 in Claude Code and this
session stops at $20 of extra usage, counted from now. At the
limit, the next prompt is blocked before it spends a
cent; resend the same prompt within 5 minutes and it
goes through, because repeating yourself is explicit consent.
The caps act only while you are in extra usage: in-plan usage
is never counted, never blocked.
One spending cap, applied per month. Nothing per session, per day or per week. When the credits run out, Fable 5 stops mid-session.
A cap on Claude Code spend per session, per day, per week or per month. The prompt is blocked before it spends, at the number you chose.
Blocked before it spends. Resend the same prompt to pass anyway. You stay in charge.
Anthropic controls as documented in Anthropic’s own support articles on extra usage and spending limits, verified July 2026.
The daily gesture and the full command family live on the /tokenkarma command page; the control-by-control comparison with Anthropic's native options is on the cap page.
Frequently asked questions
What is Claude extra usage?
Extra usage (also called usage credits) lets Pro and Max users keep working past their plan’s included limits. Instead of waiting for the window to reset, your prompts are billed per token at standard API rates, on top of the subscription. It is opt-in, and Anthropic’s only spend control on it is a monthly cap.
What does Claude extra usage cost?
Standard API rates per model. The number that matters since July 7, 2026 is Fable 5: $10 per million input tokens and $50 per million output tokens, exactly twice the Opus 4.8 rate, and it tokenizes roughly 30% more tokens for the same content. Cheaper models bill proportionally less, which is why the model mix decides the bill.
Can I limit what extra usage spends?
Anthropic offers one monthly spending cap, account-wide. tokenkarma caps Claude Code extra usage per session, per day, per week or per month, and blocks the prompt before it spends. The caps act only while you are in extra usage; usage included in your plan is never counted or blocked.
How do I see extra usage happening live?
tokenkarma shows it as it burns: a red extra-usage alert in the Mac menu bar island, a live burn ticker, and an overage card on the dashboard with your monthly figure. You see the spend while it happens, not on the invoice.
Does this cover every Claude surface?
The live tracking covers your Claude account. The caps apply to Claude Code specifically, through the guard the Mac app installs opt-in. We do not claim to block spend on other Claude surfaces.
Extra usage without the invoice surprise
tokenkarma shows the extra-usage burn live in your menu bar and blocks Claude Code spend at the number you chose.
Works in your browser and on your Mac.
Not affiliated with Anthropic. Extra usage mechanics and controls reflect Anthropic's support documentation as of July 2026 and can change. Caps apply to Claude Code via the guard, a Mac app opt-in.