The Claude Cost Optimisation Workbook

What it is

The Cost Optimisation Workbook is a downloadable PDF built directly from this site's own facts module — the same source every calculator on ClaudeHow reads from — rather than written separately as marketing copy. It exists for the reader who wants the reasoning behind the calculators laid out once, in order, rather than re-deriving it tool by tool: which lever (model choice, prompt caching, batching, or effort level) matters most for a given usage shape, in what order to apply them, and the specific numbers each lever moves.

Who it's for

Anyone shipping a production integration against the Claude API who wants a single reference document to hand to a teammate, rather than pointing them at six separate calculator pages and asking them to piece the reasoning together themselves. It assumes you already know roughly how your application calls Claude — it is not an introduction to the API.

How it's kept accurate

Because the workbook is generated from the same facts module the live site reads from, it carries the same discipline: every figure in it traces back to a dated, sourced entry, and it is regenerated whenever the underlying facts module changes materially — a pricing update, a new cache-minimum table entry, a model generation change. A workbook downloaded today reflects the facts module as of its own stated generation date, printed on its cover page; it is not a living document, so treat its date the same way you'd treat any other cost figure on this site — current as of that date, and worth re-checking against the live tools if meaningful time has passed.

Exactly what's in the download

This is written from the generated file itself, not from a plan — every count below was checked by opening the PDF. It's a 6-page, 10-section workbook: the four-lever framework (which order to apply model choice, prompt caching, batching and effort level, and why that order matters); a full current pricing table covering every model this site prices; the minimum-cacheable-prefix table grouped by floor; a prompt-caching economics table (read/write multipliers and the break-even request count for both TTLs); a batch-API economics table; a five-level effort guide (low through max, what each is for and the risk of choosing it wrong); the tiktoken-undercount correction; and three original worked scenarios — an agentic coding assistant, a document-heavy RAG assistant, and a nightly batch classification job — each priced end to end and not published anywhere else on this site.

What it does not do

It does not replace the live calculators — it's the reasoning behind them laid out once, in order, to hand to a teammate. And it does not invent a subscription plan price any more than the rest of the site does: the RAG scenario above prices the API side and asks you to compare it against what your own plan actually costs, exactly like the live subscription-vs-API calculator.

Price and delivery

The workbook is a one-time purchase, delivered as an immediate PDF download after payment through Stripe. We never see or store your card details — Stripe handles payment entirely. Because it is a digital product delivered immediately, the sale is generally final; if something goes wrong with your download or payment, contact us and we'll make it right.

Not ready to buy yet?

Every number the workbook is built from is already free to use across this site's calculators and reference pages — buying it is a way to support the ongoing maintenance this site requires, not a requirement for getting the underlying numbers right. If you'd rather support the site without a static document, /support-us/ also takes a direct one-time contribution, and either route funds the same weekly re-verification work this site depends on to stay accurate.