Real-time directory structure for Cursor AI. Your AI always knows your codebase's structure.

The AI in Cursor only gets a high-level overview of your project:

apps/ web/ src/ app/ hooks/ constants/ 

The AI can't see individual files, subdirectories, or what's actually inside those folders. It knows you have a hooks/ directory but not what hooks exist.

Twiggy fixes this by generating a detailed, real-time directory tree that shows every file and folder. Now the AI can see:

• All your individual hook files
• Nested directory structures
• Every component, utility, and config file
• The complete picture of your codebase

pip install -e .

twiggy init # setup in any project (interactive) twiggy init --defaults # setup with defaults (no questions) twiggy watch # start watching for changes twiggy stats # show index size + estimate

Creates .cursor/rules/file-structure.mdc that auto-updates when you add/remove files.

Ignores the usual stuff: node_modules, __pycache__, .git, dist, etc.

If you want the codebase index to show compact export names instead of full signatures and members, set this in twiggy.yml:

indexing: detailLevel: compact

Why track twiggy.yml but not the generated rule file?

The Problem: If we tracked the generated .cursor/rules/file-structure.mdc file, it would claim to be "updated in real-time" even when you're not running Twiggy. This creates a lying, stale file that misleads the AI about the current structure.

The Solution:

• Track twiggy.yml (config) - tells the team "this project uses Twiggy"
• Don't track .cursor/rules/file-structure.mdc (generated output) - only exists when you're actively running the watcher
• If you see the rule file, you know it's actually real-time and trustworthy for your session
• If you don't see it, you haven't set up Twiggy yet (and the AI won't get false promises)

This way the rule file is either accurate or doesn't exist - never stale.