Focus Fold
AI-powered code folding for VS Code. Focus Fold uses LLMs to automatically identify your core business logic and fold away boilerplate — so you can read what actually matters.
Install
From the Marketplace (recommended):
- Open VS Code
- Press
Ctrl+P / Cmd+P
- Paste:
ext install YiSays.focus-fold
Or search "Focus Fold" in the Extensions panel.
How It Works
As projects grow, clean code gets buried under imports, error handling, logging, and validation. Focus Fold sends your code to an LLM that identifies which regions are core business logic and which are boilerplate — then folds the boilerplate away automatically.
project-root/
.fold/
src/
main.py.json ← fold data (gitignore this, or commit it)
src/
main.py ← your source, now nicely focused
Features
| Feature |
Description |
| 🤖 AI Analysis |
LLM identifies core logic vs. boilerplate across your codebase |
| 🏷️ Smart Labels |
Click-to-toggle labels above each foldable region |
| 📊 Status Bar |
Live fold state indicator with region count |
| 🔄 Auto-Fold on Open |
Files with fold data auto-collapse boilerplate when opened |
| 💾 Auto-Reanalysis |
Re-analyzes on save to keep folding current |
| 📁 Centralized Storage |
All fold data in .fold/ — no clutter in source directories |
| 📝 FOCUS.md |
Auto-generated codebase overview with clickable navigation |
| 🔌 Multiple LLM Providers |
z.ai (GLM-4.7), OpenAI (GPT-4), OpenRouter (Claude, Llama, etc.) |
Setup
Open Settings (Cmd/Ctrl + ,) and search for "Focus Fold":
| Setting |
Description |
focusFold.llmProvider |
Choose: zai, openai, or openrouter |
focusFold.zaiApiKey |
z.ai API key |
focusFold.openaiApiKey |
OpenAI API key |
focusFold.openrouterApiKey |
OpenRouter API key |
focusFold.openrouterModel |
Model for OpenRouter (default: anthropic/claude-3.5-sonnet) |
2. Get an API Key
z.ai (GLM-4.7) — recommended for cost efficiency:
- Visit bigmodel.cn → sign up → copy your key
- Endpoint:
https://api.z.ai/api/paas/v4/ · Model: glm-4.7
OpenAI: platform.openai.com → API Keys
OpenRouter: openrouter.ai → Keys (pay-as-you-go, access to 100+ models)
3. Analyze Your Code
Open a Python file and run from the Command Palette (Cmd/Ctrl + Shift + P):
Focus Fold: Analyze Current File
Boilerplate regions will be folded immediately. Smart labels appear above each region.
Commands
| Command |
Description |
Focus Fold: Analyze Current File |
Analyze the open Python file |
Focus Fold: Analyze Python Project |
Analyze all Python files in the workspace |
Focus Fold: Re-analyze Current File |
Force re-analysis of the current file |
Focus Fold: Test API Connection |
Verify your API key works |
Focus Fold: Fold All |
Fold all regions in the current file |
Focus Fold: Unfold All |
Unfold all regions |
Focus Fold: Refold to Defaults |
Reset to default fold state |
Focus Fold: Toggle Imports |
Toggle imports visibility |
Focus Fold: Toggle Error Handling |
Toggle error handling visibility |
Focus Fold: Toggle Logging |
Toggle logging visibility |
Focus Fold: Toggle All Boilerplate |
Toggle all non-business-logic regions |
Focus Fold: Generate FOCUS.md |
Generate project-wide codebase overview |
Focus Fold: Open FOCUS.md |
Open the FOCUS.md file |
Focus Fold: Migrate to .fold Folder |
Migrate old sidecar .fold.json files |
Focus Fold: Show Folding Information |
Show fold info for the current file |
Category Classification
| Category |
Default |
Description |
business-logic |
Visible |
Core algorithms and main logic |
imports |
Folded |
Import statements |
error-handling |
Folded |
Try-except blocks, error checking |
logging |
Folded |
Log statements, debug output |
helper-functions |
Folded |
Utility functions |
data-loading |
Folded |
File I/O, data fetching |
validation |
Folded |
Input validation |
tests |
Folded |
Test code |
configuration |
Folded |
Constants and settings |
Status Bar & Smart Labels
Status bar shows current fold state:
- ○ ? — Not analyzed (click to analyze)
- › 7 — Folded (7 regions, click to unfold all)
- ∨ 7 — Unfolded (click to fold all)
Smart labels appear above each foldable region. Configure the style:
| Style |
Example |
descriptive |
📦 Import requests and pandas |
compact |
📦 13 |
minimal |
[13] |
All Settings
| Setting |
Default |
Description |
focusFold.llmProvider |
zai |
LLM provider |
focusFold.autoAnalyzeOnSave |
true |
Re-analyze on save (only if already analyzed) |
focusFold.promptOnOpen |
true |
Prompt to analyze when opening unanalyzed files |
focusFold.showSummaryInStatus |
true |
Show file summary in status bar |
focusFold.autoUpdateFocus |
true |
Auto-update FOCUS.md when files are analyzed |
focusFold.smartFoldLabels |
true |
Show smart labels on folded regions |
focusFold.showFoldLabels |
true |
Show labels above folded regions (CodeLens) |
focusFold.foldLabelStyle |
descriptive |
Label style: compact, descriptive, minimal |
focusFold.showFoldCountInStatus |
true |
Show fold region counts in status bar |
focusFold.autoFoldOnSave |
true |
Auto-apply folding when saving analyzed files |
focusFold.enableContextCache |
true |
Cache project context for faster re-analysis |
focusFold.contextCacheMaxAge |
5 |
Cache max age in minutes |
focusFold.contextCacheMaxEntries |
10 |
Max cached project contexts |
focusFold.useGlobalContext |
true |
Use global project context for better accuracy |
FOCUS.md — Codebase Overview
FOCUS.md is an auto-generated markdown file providing a high-level overview of your analyzed codebase:
- Project statistics (files, core regions, boilerplate regions)
- Per-file summaries with descriptions
- Clickable VS Code links to open files at specific line numbers
- Grouped by directory
Generate via: Focus Fold: Generate FOCUS.md
Migration from v0.1.x
Version 0.2.0+ uses a centralized .fold/ folder. If you have old sidecar .fold.json files:
Focus Fold: Migrate to .fold Folder
Recommended .gitignore entry
.fold/
FOCUS.md
Roadmap
- [ ] JavaScript / TypeScript support
- [ ] Java and Go support
- [ ] Manual region editing
- [ ] Custom category preferences
- [ ] Folding annotations in gutter
See BLUEPRINT.md for the full vision.
Contributing
See CONTRIBUTING.md for development setup, debug workflow, and contribution guidelines.
License
MIT © 2026 YiSays
Acknowledgments
