Tested guides and free scripts on AI token usage and API costs. https://neverpaytwice.pages.dev/ 2026-11-15T12:00:00Z NeverPayTwice https://neverpaytwice.pages.dev/claude-thinking-tokens-cost 2026-11-15T12:00:00Z

Claude's thinking is billed as output tokens, even when you can't see it. How to see how much of your bill is thinking, which models let you turn it off, how the effort setting lowers it, and two mistakes that make it cost more.

https://neverpaytwice.pages.dev/claude-batch-api-cost 2026-11-15T12:00:00Z

Anthropic's Message Batches API charges 50% of the standard price on every token, cache included. Batch prices for each Claude model, what fits in a batch and what doesn't, the limits, and a free script that shows what batching would save on your own requests.

https://neverpaytwice.pages.dev/claude-api-bill-too-high 2026-11-15T12:00:00Z

Claude API bill higher than you expected? Hidden thinking tokens, a prompt cache that never hits, resent history, fast mode and the wrong model are the usual causes. How to check each one, and a free script that shows where your money went.

https://neverpaytwice.pages.dev/claude-code-subscription-vs-api 2026-11-14T12:00:00Z

Is a Claude Pro, Max or Ultra plan cheaper than paying per token for Claude Code? Price a month of your own usage at API rates and compare. Plan prices, the two counting mistakes that skew it, and a tested Python script.

https://neverpaytwice.pages.dev/claude-prompt-caching-cost 2026-11-14T12:00:00Z

Claude prompt caching makes cache writes cost more and cache reads cost far less. When it pays off, when it quietly costs you extra, and a tested Python script that checks your own usage logs.

https://neverpaytwice.pages.dev/duplicate-ai-api-charges 2026-11-14T12:00:00Z

Retries, double-clicks and re-run pipelines quietly buy the same AI answer again. Here is how to find requests you paid for more than once, with a tested Python script, and what to do about each kind.

https://neverpaytwice.pages.dev/openai-api-cost-per-request 2026-11-14T12:00:00Z

OpenAI's Usage API reports totals per time bucket, not the cost of each request. Here is how to work out the cost of every request from its usage block, for Chat Completions and the Responses API, with a tested Python function.

https://neverpaytwice.pages.dev/anthropic-api-cost-per-request 2026-11-14T12:00:00Z

Anthropic's usage reports show totals, not the cost of each request. Here is how to work out what every Claude API request cost from its usage block, including cache writes, cache reads and thinking tokens, with a tested Python function.

https://neverpaytwice.pages.dev/claude-code-token-count 2026-11-13T12:00:00Z

Adding up Claude Code's usage logs can give you nearly double your real token usage. Here is why the logs repeat themselves, how to count correctly, and a free tool and script to do it.