What does your CLAUDE.md cost you?
Claude Code loads your project's CLAUDE.md into context at the start of every session. Paste yours below for an estimate of its size and what that read costs on the API. Nothing leaves your browser - this page has no server and no analytics on your text.
Runs entirely in this tab. Close it and your text is gone.
Your longest lines
The ten longest lines, by estimated tokens. Long prose lines are candidates for moving into a separate doc the agent opens only when needed.
| # | line | est. tokens | preview |
|---|
What this estimate does and doesn't show
Token estimate: characters divided by 4, a rough average for English and code. The real count varies with content and tokenizer - Anthropic notes that models from Claude 4.7 on tokenize the same text into about 30% more tokens. Treat every number on this page as a rough estimate, not an exact count. One "session" here means the file is loaded into context once; real sessions add conversation, tool output, and re-reads on top. Cost uses the first-party API list price you selected (source: docs.claude.com/en/docs/about-claude/pricing, checked Nov 13, 2026), uncached, per million input tokens. Prompt caching makes repeat reads far cheaper (the cache-hit figure above). If you use a flat-rate plan like Pro or Max, tokens are not metered per request - for you this file costs context-window room, not money. Nothing here measures whether the instructions work, only what they weigh.