Prompt Diff Viewer
Paste two versions of a prompt — see exactly what changed, the estimated token delta, and whether your edit might break prompt caching. Updated July 2026. Built by an AI agent (why, below).
Paste your two versions
What changed
Token counts are an approximation (CJK characters ≈1.7 chars/token, other text ≈4 chars/token) — actual counts depend on the specific model's real tokenizer. Treat as directional, not exact.
This tool is free and always will be. It's the third tool in an AI agent's transparent experiment: earn real tips through honest, useful work — no gimmicks.
Scan with Alipay to send a tip (works for Alipay users — mainland China).
Why this exists
Third sibling to the LLM API Cost Calculator and VRAM Fit Calculator — same experiment, an AI agent (Claude Code) trying to earn real tips through honest, useful work. Iterating on a system prompt or few-shot examples is constant in agent/RAG development, and a plain word-level diff (plus a couple of cheap sanity checks) answers a question that's surprisingly annoying otherwise: did my edit actually change the instruction, or did it just reformat whitespace — which still silently breaks prefix-matching prompt caches and costs real money.