vague
messy
half-formed
structured
on-topic
concise
Prompy rewrites your rough draft into a clear, well-structured prompt — right where you type, in ChatGPT, Claude, Gemini and more. Bring your own key or run a fully local model.
Free. You bring the API key. Works in your browser; desktop app on the way.
Write your prompt however it comes out. Messy and half-formed is fine.
Prompy streams a sharper, better-structured version into a quiet card beside your cursor.
One click swaps it in. Or copy it, switch styles, generate another variation.
PROMPY
thinking
A fuzzy prompt usually takes a few clarifying rounds and a redo. Every round costs more tokens than getting it right the first time.
Illustrative estimate: assumes a vague prompt needs ~3 follow-up rounds. Actual savings vary by task and model.
Claude, GPT, Gemini, Mistral, or fully local with Ollama or LM Studio. Bring your own key, swap providers anytime, or skip the cloud entirely. Prompy doesn't care which brain does the rewriting.
No accounts, no analytics, no middleman server. Your key lives on your device, encrypted at rest on desktop, and every request goes provider-direct. What you type is between you and your model.
Detailed, Concise, Professional, or Coding. Switch and re-run from the card.
Reads the on-page chat so follow-ups fit what you're actually discussing.
Not sure? Generate a few alternatives and pick your favorite.
Improve the focused field with a keystroke, or right-click any text.
Suggestions appear token-by-token, with caching so repeats are instant.
Prompy has no backend and no database, so there's nothing to breach. Your prompts go straight to the provider you chose (or a model running on your own machine), authenticated with your own key.
prompy server doesn't exist — we never see your draft
The same rough idea — “make a prompt to help me write cold emails for my saas” — rewritten in each style.
ChatGPT, Claude, Gemini and the other usual suspects are built in. But new AI tools land every week, and no list stays current.
So instead of a list, there's a shortcut. Focus any prompt field, press the keys, and Prompy wakes up for the rest of the tab — same card, same styles. It loads only when you call it and reads nothing in the background.
on Windows / Linux: Ctrl + Shift + P
Point Prompy at a model running on your own machine (Ollama or LM Studio) and you don't need an API key, a network connection, or anyone's permission. Nothing else changes. Just zero dollars and zero packets leaving the laptop.
People reach for this when they're working with health records, customer data, NDA material, or anything where "this prompt is going to a third-party API" isn't a sentence they're willing to say. Also when they're on a plane.
Two minutes, three lines:
# 1. Install Ollama (ollama.com), then pull whatever you like ollama pull llama3.2 # 2. In Prompy, pick "Local" and use this URL http://localhost:11434/v1 # 3. Browser extension only — let it talk to Ollama OLLAMA_ORIGINS=chrome-extension://* ollama serve
LM Studio works the same way at http://localhost:1234/v1 (CORS already on).
The tutorial has the longer version →
Live on the Chrome Web Store — free to use, bring your own key.