Compress your AI prompts directly inside VS Code — before sending to GitHub Copilot, ChatGPT, Claude, or any LLM.
Why
Every word you send to an AI costs tokens. Long prompts = slow replies + higher API bills. TokenLean strips bloated words, filler phrases, and verbose patterns — same meaning, fewer tokens.
Commands
| Command | Shortcut | Action |
|---|---|---|
| Compress Prompt | Cmd+Shift+L |
Compress selected text |
| Compress + Directives | Cmd+Shift+Alt+L |
Compress + add output rules |
| Add Custom Swap | Command Palette | Add your own word swap |
| View / Manage Swaps | Command Palette | See and delete your swaps |
| Refresh Swaps | Command Palette | Pull latest swaps from cloud |
How to use
- Select any prompt text in the editor
- Press
Cmd+Shift+Lor right-click → TokenLean: Compress Prompt - Choose Replace or Copy to Clipboard
- Paste into Copilot Chat or any AI tool
What it compresses
Word swaps (70+ built-in)
implementation→buildauthentication→authdocumentation→docsutilize→usecurrently→nowsufficient→enough
Phrase stripping
Could you please→ removedI would like to know→tell meDue to the fact that→becauseIn order to→to
Output directives Tell the LLM exactly how to reply — format, length, no fluff.
Personal dictionary
Add your own domain-specific swaps via TokenLean: Add Custom Swap. Stored locally in VS Code. Applies automatically on every compression.
Privacy
All compression runs locally. No data sent to any server except fetching the shared swap list from Supabase (read-only, anonymous).
Links
- Web app: https://tokenlean.lovable.app
- Chrome extension: Chrome Web Store → search TokenLean