I kept running into the same friction loop: tweak a prompt, spin up the project, wait for deps, re-run the script, get an error, try again. Each cycle was 8+ minutes. At 20 iterations a day that's a real chunk of time gone before I've learned anything useful.

  So I built PromptFast — a browser-based prompt playground that skips all
  of that. You open it, paste a prompt, and run it against whichever model
  you want in under 5 seconds.

  The features I use constantly:

  - 13+ models across OpenAI, Anthropic, and Google — switch in one click
  - Dynamic variables ({{tone}}, {{recipient}}) so you can test multiple
    scenarios without rewriting
  - Side-by-side model comparison — great for deciding between GPT-4o and
    Sonnet on a specific task
  - Token and cost breakdown per run — helps avoid surprise bills
  - File upload context (PDF, DOCX, CSV, XLSX, JSON) — useful for
    document-heavy prompts
  - System prompt field + prompt history so you can reload and iterate
    on past runs

  Your API keys are encrypted client-side (AES-GCM) and never stored on
  my server. The backend is FastAPI + LangGraph; the frontend is Next.js.

  It's a paid tool (€5/mo, or €45/yr, or €95 lifetime) — no free tier
  currently, though you can try the demo on the landing page. Still early
  days and happy to hear what's missing or broken.

  https://promptfast.app