Skip to content
← Projects
2026·Active

savedyouatoken

An open-source tool that audits LLM system prompts for waste, prices each finding in real dollars per month, and returns a lossless rewrite — no account, no upload, no model call.

  • TypeScript
  • Next.js 16
  • React 19
  • Tailwind CSS 4
  • Vitest
  • Vercel

savedyouatoken analyses a system prompt and returns a ranked list of what's wasting tokens — each finding priced in dollars per month at your real request volume — plus a rewritten version you can copy directly. The whole analysis runs client-side: no account, no upload, no model call, which is also why it costs nothing to use.

The core engine is dependency-free TypeScript shared verbatim across the web app, CLI, and test suite, with a tokenizer injected per model family rather than assumed. The web analyser is a fully static Next.js 16 app on Vercel; a companion CLI and runtime SDK apply the same rules to files on disk and to live requests. Open source under MIT.