Skip to content
1 month freeJoin the UsageMeter Discord and get 1 month of Pro freeJoin the Discord

Free tool

What would Claude Code cost on the API?

Paste your /usage output or a ccusage export: everything is computed in your browser, nothing is sent.

Paste the block that starts with “Total cost:” or a JSON from ccusage.

Cache writes priced at

Calculated in your browser. Nothing you type or paste is sent anywhere.

Where the numbers come from

Claude Code counts four kinds of tokens for each model: input, output, cache reads and cache writes. The Session block of /usage (/cost is the same command) lists them per model, and ccusage exports the same counts from your local transcripts. The calculator multiplies each count by the public API price of that model.

/usage only sees this machine: its figures come from the local session history, without your other computers or claude.ai. Run it on each PC, or export with ccusage on each one, to get the whole picture.

Which prices

Anthropic’s public API prices per million tokens, checked on September 22, 2026. A model is matched by the longest known prefix of its id: claude-opus-5-5-20260922 is priced as claude-opus-5-5, never as claude-opus-5. A model that is not in the table is shown but never guessed.

Cache reads cost a tenth of the input price on most models (less on Claude Opus 5.5 and Claude Fable 5.1). Cache writes cost 1.25× the input price for a 5-minute lifetime, or 2× for one hour.

Why the cache changes everything

Claude Code sends the whole conversation with every request. Most of it is read back from the prompt cache, at a fraction of the input price, which is why cache reads are usually the biggest token count and a small part of the bill. The “Without the cache” line prices the same tokens as plain input to show what the cache saved.

The lifetime matters too: on a Pro or Max subscription Claude Code keeps its main cache for an hour, and for five minutes on an API key or usage credits. After a longer break the cache has expired, and the first request writes it again in full.

Subscription or API

Pro and Max do not bill per token: the dollar figure of /usage is meant for API users, and on a subscription it is not what you pay. What limits you is the 5-hour session and the weekly limit, which the dollar figure does not show. The API equivalent is still a useful yardstick of how much you use Claude Code.

Anthropic’s own figures for Claude Code billed at API rates: about $13 per developer per active day, $150 to $250 per developer per month, and under $30 per active day for 90% of users. Past your limits on Pro and Max, usage credits are billed at these API rates.

Questions

Do I pay this amount with a Claude subscription?

No. Pro and Max have a fixed price and limits: the API equivalent only tells you what the same work would have cost on an API key. You pay per token only with usage credits, once you go past your limits.

Why does my result differ from the dollar figure of /usage?

The usual reasons: cache writes priced at one hour instead of five minutes, a price that changed since your Claude Code version was released, or a model missing from our table (shown, never guessed).

What are cache reads and cache writes?

A cache write stores the start of a conversation so that the next requests can reuse it; a cache read reuses it at a fraction of the input price. Long Claude Code sessions are mostly cache reads.

How do I get the total of all my computers?

/usage and ccusage only read the machine they run on. UsageMeter Link reads Claude Code on each Windows PC and adds everything up in your UsageMeter account, per model and per day.

Sources

Checked against Anthropic’s pages on September 24, 2026. Anthropic can change these rules at any time.

UsageMeter is an independent tool, not affiliated with Anthropic. Claude and Claude Code are Anthropic’s products.