Turn your messy thoughts into perfect AI prompts — instantly, offline, and for free.
PromptForge is a lightweight, single-file HTML app that takes your raw "word vomit" and transforms it into a polished, optimized prompt for Claude or Gemini — powered entirely by a local AI model (no API keys, no subscriptions, no internet required after setup).
- 🧠 Brain dump → polished prompt in seconds
- 🤖 Claude & Gemini optimized outputs (different prompt styles for each)
- 🎨 5 tone options — Professional, Casual, Technical, Creative, Concise
- 🎙️ Voice input — speak your thoughts instead of typing
- 🕐 Prompt history — last 20 prompts saved automatically
- 🔗 Quick links to open Claude and Gemini directly
- 💻 100% offline — runs on your machine via Ollama
- 📄 Single HTML file — no install, no build step, no dependencies
If PromptForge saves you time, consider buying me a coffee!
Every coffee keeps this project going and motivates new features. Thank you! 🙏
You need Ollama installed on your machine. It's free, open source, and runs entirely locally.
Windows:
- Go to ollama.com
- Click Download for Windows
- Run the installer — Ollama will appear in your system tray when running
Mac:
- Go to ollama.com
- Click Download for Mac
- Open the
.dmgand drag Ollama to your Applications folder - Launch Ollama from Applications
Open a terminal (PowerShell on Windows, Terminal on Mac) and run:
ollama pull llama3.2This downloads the model (~2GB). You only need to do this once.
This is a one-time setup per machine so your browser can communicate with Ollama.
Windows (PowerShell):
setx OLLAMA_ORIGINS "*"Then fully quit Ollama from the system tray and relaunch it.
Mac (Terminal):
launchctl setenv OLLAMA_ORIGINS "*"Then restart Ollama.
Tip: If you ever restart your Mac, you may need to run the
launchctlcommand again before launching Ollama.
Alternative (no restart needed): Instead of the above, you can always launch Ollama from the terminal with:
# Windows PowerShell
$env:OLLAMA_ORIGINS="*"; ollama serve
# Mac Terminal
OLLAMA_ORIGINS="*" ollama serveLeave this terminal window open while using PromptForge.
- Download
promptforge.htmlfrom this repo - Open it in Google Chrome or Microsoft Edge
- Make sure Ollama is running in the background
- Start generating prompts!
- Choose your AI — Select Claude or Gemini depending on where you'll paste the prompt
- Brain dump — Type (or speak 🎙️) your raw, unpolished thoughts
- Pick a tone — Professional, Casual, Technical, Creative, or Concise
- Generate — Hit the button or press
Cmd/Ctrl + Enter - Copy & paste — Use the Copy button, then paste into Claude or Gemini
| Platform | Supported | Notes |
|---|---|---|
| Windows + Chrome | ✅ | Recommended |
| Windows + Edge | ✅ | Works great |
| Mac + Chrome | ✅ | Recommended |
| Mac + Safari | Voice input may not work | |
| Firefox | Voice input not supported |
"Cannot connect to Ollama" error:
- Make sure Ollama is running (check your system tray on Windows, menu bar on Mac)
- Make sure you set
OLLAMA_ORIGINS="*"and restarted Ollama - Try running
ollama servein a terminal with ORIGINS set inline (see Step 3)
Voice input not working:
- Use Chrome or Edge
- Make sure you allowed microphone permissions when prompted
- Check that your browser has microphone access in system settings
Blank output:
- Check that
llama3.2is fully downloaded (ollama pull llama3.2) - Open browser console (F12 → Console) and check for error messages
- Pure HTML, CSS, JavaScript — zero dependencies, zero build tools
- Ollama — local AI model runner
- llama3.2 — the underlying language model
- Web Speech API — for voice input
Contributions are welcome! Feel free to open an issue or submit a pull request.
Ideas for future features:
- Custom model selector (swap llama3.2 for any Ollama model)
- Export history as markdown
- More AI targets (Perplexity, ChatGPT, etc.)
- Prompt templates / presets
MIT License — free to use, modify, and distribute.
Built with ❤️ and a lot of word vomit.