Independent · Updated on a visible cadence
Nobody budgets for tokens. Everyone pays for them.
Most people meet Claude’s cost model twice: when the invoice arrives, and when they get cut off mid-task. This site explains both — what each plan and model actually costs, why limits hit when they do, and which habits quietly burn the most context.
The thesis, in three lines
Token spend is not a pricing problem, it is a habit problem. The same task can cost five times as much depending on how much history you drag along with it, how often you restate context, and whether you let a long thread run instead of starting a clean one.
Anthropic’s documentation tells you the rates. It does not tell you which of your own patterns are expensive. That gap is what these pages are for, and it is what the audit puts a name to in your own transcripts.
Start where your question is
- PricingEvery plan and every model price in one maintained table — Pro, Max, Team, API.Start with API pricing
- UsageWhy you got cut off, when the window resets, and what extra usage really buys.Start with usage limits
- TokensWhat a token is, how the context window fills up, and which habits waste the most.Read the token guide
- CoworkCowork, Claude Code and subagents — what each one costs you in context.Cowork vs Claude Code
- OptimizeThe playbook: concrete changes that cut spend without cutting output quality.Open the playbook
Or skip the reading and have someone look at your actual prompts
Paste your last twenty prompts and Claude’s replies into one box. You get back a written report: where the tokens went, which of your habits cost the most, and the specific rewrites that would have cut it. No call to book, no calendar to negotiate.
Written, not scheduled
A document, not a call.
Your prompts stay yours
Retention and deletion spelled out.
Numbers you can act on
Per-habit cost, not general advice.