Paste two prompt versions and see exactly what changed. Word-level diff, token delta, and cost comparison.
Token estimates are approximate (~+/-15% vs real tokenizers). Costs are input-only estimates based on public API rates in USD per million tokens.