A free calculator estimates what Claude Code costs when you pay per token
A small tool shared on Show HN turns Claude Code usage into a monthly dollar estimate, which helps anyone paying by the token set a budget first.
The short version
- The estimator covers API token spend when Claude Code calls ModelFold, not a Claude subscription.
- Three presets (Light, Regular, Heavy) set days, requests and tokens, and you can edit all of them.
- In the example shown, 7 credits per request adds up to 560 credits, or $2.80, per day.
- Credits convert at 200 credits = $1, and failed generations are not charged.
- The post had one point and no comments when listed, and we have not checked its math.
Why it matters
Coding agents can use far more text than a normal chat, so a bill can surprise you. Putting a number on it before you start lets you set a ceiling and compare it with real usage. The tool is niche, but the habit applies to any provider.
What the calculator does
As modelfold.ai reported on its own page, the tool estimates the API token spend when Claude Code calls ModelFold. It does not price a Claude subscription, which Anthropic bills separately and which is not included here.
Each request is rounded up, then multiplied by the number of requests and days. In the example on the page, 1,760 requests in a month works out to 560 credits, or $2.80, per day, and 7 credits per request.
Three presets, all editable
The page gives three planning assumptions. They are guesses to start from, not a measurement of your account.
- Light: 20 days, 30 requests, 12,000 input tokens and 800 output tokens.
- Regular: 22 days, 80 requests, 25,000 in and 1,500 out.
- Heavy: 22 days, 160 requests, 50,000 in and 3,000 out.
A token is a chunk of text, roughly a word fragment. The page says to change the counts to match the context you actually send.
Why agent bills climb
The newsletter's explanation: in an agent loop, each step resends the whole conversation so far, plus new material such as file contents and command results. Early text is paid for again at every step. Cost therefore follows history length times number of steps, not the length of the answer.
As a made-up toy case, if each step adds 5,000 tokens, 10 steps process 275,000 tokens and 20 steps process 1,050,000. Doubling the steps nearly quadruples the total. That is why long sessions can cost disproportionately more.
How to read the result
This is an estimate. Real usage varies with the size of tool output, the model and how often cached text is reused. We have not tested the calculator, and the pricing data it uses may not be current.
The tool also ties into ModelFold. To pay these prices, you point Claude Code at ModelFold's Anthropic-compatible endpoint with a ModelFold API key. That is a separate account and a separate bill from Anthropic's own.
What to watch
- Whether the calculator's estimates match real bills once people try it.
- Whether the pricing and presets stay current as providers change their rates.
Try it
Pick the preset closest to how you work, edit it to your real token counts, then compare the monthly figure with a week of actual usage on your billing page.
Sources and links
- Reportingmodelfold.ai: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by modelfold.ai · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.