context-lens

Paste a Claude prompt — see how it tokenizes, where caching boundaries live, and what each call will cost before you ship.

System prompt~123 tokens (heuristic)

User message~76 tokens (heuristic)

Model:Est. output:tokensAssume input is cache-read

Showing heuristic (~3.7 chars/token, ×1.3 for this model's newer tokenizer). Use the API for the real number.