A lightweight, simplified version of Plandex that orchestrates multiple AI agents to handle complex development tasks. Each agent specializes in a specific role, working together to break down requests, write code, review outputs, and provide architectural guidance.

• Multi-Agent Architecture: Five specialized AI agents working in harmony
• Plug & Play Setup: Simple Ollama integration - just provide an endpoint
• Role-Based Task Distribution: Each agent focuses on what they do best
• TypeScript Implementation: Type-safe, modern JavaScript development
• PDF Export: Built-in markdown-to-PDF conversion capabilities
• Lightweight & Fast: Minimal overhead, maximum efficiency
• Local AI Support: Works with locally hosted models via Ollama

"You are a Planner. Break down user requests into clear, actionable steps."

Analyzes complex requests and creates structured, step-by-step execution plans.

"You are a Coder. Write clean, working code that solves the described tasks."

Implements solutions with clean, maintainable code following best practices.

"You are a Summarizer. Summarize long outputs into concise points."

Distills lengthy outputs into clear, actionable summaries and key insights.

"You are an Architect. Propose high-level design choices and trade-offs."

Provides system design recommendations, architectural patterns, and technical trade-offs.

"You are a Reviewer. Check outputs for errors, improvements, and clarity."

Quality assurance specialist that validates outputs for correctness and suggests improvements.

• Ollama installed and running
• A compatible language model (tested with Qwen2.5:7b)
• Node.js 16+ and npm/yarn
• TypeScript (for development)

# Visit https://ollama.ai/ for installation instructions
# Or use curl (Linux/macOS)
curl -fsSL https://ollama.ai/install.sh | sh

# Pull Qwen2.5 (recommended)
ollama pull qwen2.5:7b

# Or try other models
ollama pull llama2
ollama pull codellama

ollama serve

Set your Ollama endpoint in the configuration:

# Default Ollama endpoint
export OLLAMA_ENDPOINT="http://localhost:11434"

git clone https://github.com/Magnus0969/Plandex-lite.git
cd Plandex-lite

# Install dependencies
npm install

# Build TypeScript
npm run build

# Run the application
npm start
# or
node plandex.js

The system is designed to be plug and play with any Ollama endpoint. Update your config_models.json:

{
  "ollama_endpoint": "http://localhost:11434",
  "model": "qwen2.5:7b",
  "timeout": 300,
  "max_tokens": 4096,
  "temperature": 0.7
}

While tested primarily with Qwen3:4b, Plandex-lite works with various Ollama-compatible models:

• ✅ Qwen3:4b (Recommended)
• ✅ Llama 2/3
• ✅ Code Llama
• ✅ Mistral
• ✅ Gemma
• ✅ Any Ollama-compatible model

Request: "Create a todo app with React and Node.js"

🎯 Planner: Breaks down into frontend/backend tasks
💻 Coder: Implements React components and Express API
🏗️ Architect: Suggests project structure and data flow
🔍 Reviewer: Validates code quality and suggests improvements
📝 Summarizer: Provides implementation summary and next steps

Request: "Refactor this legacy JavaScript code for better performance"

🎯 Planner: Identifies refactoring opportunities
🏗️ Architect: Proposes modern patterns and optimizations  
💻 Coder: Implements refactored code
🔍 Reviewer: Compares performance and validates improvements
📝 Summarizer: Highlights key changes and benefits

plandex-lite/
├── config_models.json     # Model configuration settings
├── core.ts               # Core application logic
├── index.ts              # Main entry point
├── markdown-to-pdf.ts    # PDF generation utilities
├── package-lock.json     # Dependency lock file
├── package.json          # Project dependencies and scripts
├── plandex.js           # Main orchestrator
├── roles.ts             # Agent role definitions and prompts
├── tsconfig.json        # TypeScript configuration
└── types.ts             # TypeScript type definitions

1. Define new role in roles.ts with a specific prompt
2. Update type definitions in types.ts
3. Implement agent logic in core.ts
4. Update configuration in config_models.json
5. Rebuild with npm run build

1. Fork the repository
2. Create a feature branch: git checkout -b feature-name
3. Commit changes: git commit -am 'Add feature'
4. Push to branch: git push origin feature-name
5. Submit a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

• Inspired by the original Plandex project
• Built with Ollama for local AI model hosting
• Tested extensively with Qwen2.5 models

• 🐛 Report Issues
• 💡 Feature Requests
• 📚 Documentation

Ready to orchestrate your development workflow? 🚀

Just point Plandex-lite to your Ollama endpoint and watch the agents collaborate to tackle your most complex development challenges!