Improve your AI prompts before you send them.
Works in VS Code, Cursor, Windsurf, Claude Code — any editor. No setup required.
You write a rough prompt. PromptPilot rewrites it into a sharper, better-structured version — then you send it straight to your AI chat in one click.
Before:
make the login faster and fix the bug in auth
After (Structure as task preset):
## Goal
Optimize login performance and fix the authentication bug.
## Context
The current login flow has a noticeable delay and an intermittent auth failure.
## Constraints
- Do not change the public API surface
- Maintain backward compatibility with existing sessions
- Add error handling for all auth failure cases
## Expected output
- Updated auth service with the performance fix applied
- Unit test covering the previously failing case
- Summary of what changed and why
- 4 improvement presets — Structure as Task / More Specific / Shorter / Add Constraints
- Works everywhere — VS Code, Cursor, Windsurf, any VS Code fork
- Zero setup — just install and use. No API key required by default.
- Your own key — optionally add an OpenAI, Groq, Anthropic, or Ollama key for unlimited use
- Send to chat — injects the improved prompt directly into Copilot Chat, Cursor Composer, or your clipboard
- Keyboard shortcut —
Ctrl+Alt+P/Cmd+Alt+P
PromptPilot uses a 3-tier LLM strategy — automatic, no configuration needed:
1. VS Code built-in LM (if you have Copilot — zero setup, free)
↓ not available?
2. Your own API key (if you added one — unlimited, your cost)
↓ no key?
3. Free hosted proxy → Cloudflare Workers AI / Llama 3.2 (always works, 30 req/day free)
Install from the VS Code Marketplace or Open VSX (search PromptPilot) or:
code --install-extension promptpilotOpen the command palette (Ctrl+Shift+P) and run:
PromptPilot: Set API Key
Supported providers:
| Provider | Model used by default | Notes |
|---|---|---|
| OpenAI | gpt-4o-mini |
Best quality |
| Groq | llama-3.1-8b-instant |
Fastest, has free tier |
| Anthropic | claude-3-5-haiku-latest |
Great for prompt tasks |
| Ollama | llama3.2 |
Local, 100% private, free |
| Custom | your choice | Any OpenAI-compatible URL |
Your key is stored securely in the OS keychain (VS Code SecretStorage — never in settings.json).
To remove your key:
PromptPilot: Remove API Key
| Setting | Default | Description |
|---|---|---|
promptImprover.userProvider |
openai |
Your API provider |
promptImprover.userModel |
(provider default) | Model name override |
promptImprover.userBaseUrl |
(provider default) | Custom base URL (Ollama, LM Studio, etc.) |
promptImprover.proxyUrl |
(built-in) | Self-host the proxy (advanced) |
The free proxy is a Cloudflare Worker that uses Cloudflare Workers AI (@cf/meta/llama-3.2-3b-instruct). You can deploy your own copy:
- Clone this repo
cd worker && npm install -g wrangler- Create a KV namespace:
wrangler kv namespace create RATE_LIMIT - Update
worker/wrangler.tomlwith your KV namespace ID - Deploy:
wrangler deploy - Set your worker URL in VS Code settings:
promptImprover.proxyUrl
git clone https://github.com/farzinhamzehi/promptpilot
cd promptpilot
npm install
npm run build # one-shot build
npm run watch # watch mode
# Press F5 in VS Code to launch the Extension Development Host- Your prompts are sent to the improvement backend (either your own key's API, or the free proxy).
- The free proxy does not log prompt content — only the IP address for rate limiting (reset daily).
- If privacy is critical, add an Ollama key to keep everything local.
See CHANGELOG.md for full version history.
- v0.1.2: Added screenshot preview, custom market & sidebar icons, Cloudflare Workers AI backend.
- v0.1.1: Added custom activity bar icon & license.
- v0.1.0: Initial release with 4 presets & 3-tier LLM engine.
MIT