🌾 Pliny Framework — Cursor Extension
✨ Features🏗️ Document Entire Repositories — Enter a GitHub Enterprise URL, Pliny clones it, analyzes the codebase, and generates full project documentation 📁 Document Open Folders — Right-click any folder or document your current workspace with one command 📄 HARVEST Single Files — Run the 7-phase HARVEST loop on any open file for deep documentation 🔍 Standards Checking — Validate code against your team's coding standards with violations shown in the Problems panel 📊 Quality Scoring — Every run scores documentation on completeness, accuracy, and consistency (target: 95%+) 🤖 Multi-Provider AI — Works with Cursor's built-in AI, OpenRouter, or Anthropic API 🎨 Sidebar Panel — Quick-access actions, recent run history with quality bars, and settings at a glance 🚀 Installation in CursorStep 1: Get the
|
| Setting | Default | Description |
|---|---|---|
pliny.provider |
auto |
LLM provider: auto, openrouter, anthropic, cursor |
pliny.model |
(empty) | Model override (e.g., anthropic/claude-3.5-sonnet) |
pliny.qualityThreshold |
0.95 |
Minimum quality score (0.5–1.0) |
pliny.maxIterations |
5 |
Max HARVEST loop iterations (1–10) |
pliny.maxFiles |
50 |
Max files to analyze per repo (10–200) |
pliny.outputDirectory |
./output |
Where generated docs are saved |
🔄 How HARVEST Works
The HARVEST engine runs 7 phases in an iterative loop until quality threshold is met:
Phase 1: 🌾 Harvest → Extract all information from the code
Phase 2: 🔬 Analyze → Identify patterns and architecture
Phase 3: 📐 Restructure → Apply 3-layer documentation template
Phase 4: ✅ Verify → Score quality (completeness, accuracy, consistency)
Phase 5: 📚 Extend → Add examples, getting started, troubleshooting
Phase 6: 🔗 Synthesize → Create cross-references and glossary
Phase 7: ✨ Transform → Polish to publication-ready Markdown
If quality < threshold → repeat all 7 phases (up to max iterations).
📄 Output Structure
Generated docs follow a 3-layer framework:
- Layer 1: Project Understanding — Purpose, architecture, tech stack, key concepts
- Layer 2: System Intelligence — Module relationships, data flows, critical paths, error handling
- Layer 3: Reference — Module inventory, API surface, config reference, dependencies, known issues
Plus: getting started guide, development workflow, troubleshooting, glossary, and 10+ cross-references per section.
🛠️ CLI Usage (Alternative)
The Pliny Framework also works as a standalone CLI:
cd pliny-framework
npm install
# Document a repo
node bin/pliny.js document https://ghe.company.com/org/repo
# Document a local folder
node bin/pliny.js document ./path/to/project
# HARVEST a single file
node bin/pliny.js harvest path/to/file.js
# Check standards
node bin/pliny.js check standards --type js path/to/file.js
🐛 Troubleshooting
❓ "No LLM provider available"
→ Run Pliny: Configure API Key to set up OpenRouter or Anthropic, or switch to Cursor AI in settings.
❓ Extension not appearing after install
→ Reload Cursor: Ctrl+Shift+P → Developer: Reload Window
❓ Clone fails for private repo
→ Make sure your Git credentials are configured for GHE: git config --global credential.helper manager
❓ Documentation quality is low
→ Increase pliny.maxIterations (default 5) and pliny.maxFiles (default 50) in settings.
❓ Output channel logs
→ Open the Output panel (Ctrl+Shift+U) and select Pliny from the dropdown to see detailed phase logs.
📋 Requirements
- Cursor v1.93+ (or VS Code v1.93+)
- Node.js 18+ (bundled with Cursor)
- Git (for cloning repositories)
- One of: Cursor Pro subscription, OpenRouter API key, or Anthropic API key
👨💻 Author
Chase Logan — Built with the Pliny R-I-S-E, C-A-R-E, and HARVEST frameworks.
📄 License
MIT