Compress LLM prompts locally — no API calls, no inference cost.
Select text in the editor, run H47: Optimize Selected Text, and get a shorter version ready to paste into Claude, ChatGPT, or Cursor.
Features
- Local & deterministic — runs in-process, ~1–15 ms typical
- Context menu — right-click selected text → H47 commands
- Configurable — compression level, target AI, token budget in Settings
- Keyboard shortcuts —
Ctrl+Shift+O(optimize),Ctrl+Shift+Alt+O(replace)
Commands
| Command | Action |
|---|---|
| H47: Optimize Selected Text | Preview result in new markdown tab |
| H47: Optimize and Replace Selection | Replace selection in place |
| H47: Optimize for Claude | Claude-optimized output, replaces selection |
| H47: Show Optimizer Info | Version and supported models |
Settings
Search h47 in VS Code Settings:
h47.compressionLevel—conservative|balanced|aggressiveh47.targetAI—generic|claude|gpt|cursorh47.maxTokens— output token budget (default 2000)h47.showPreview— preview vs replace for main optimize command
Limitations
Extractive compression works best on long, repetitive text (logs, code, conversations). Short selections may not compress. See limitations.
Also works in Cursor
Install the .vsix file: Extensions → … → Install from VSIX.
License
MIT · Copyright © 2024-2026 H47 Team