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
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.