Universal AI Software House

AI-powered software house team with permanent memory and 20+ specialized agents. Works with ANY coding tool.

Create a team of specialized AI agents for any project (new or existing) in minutes. The orchestrator analyzes your project, asks intelligent questions, generates a development plan, builds a specialized team, and executes tasks in parallel.

What's Included

• Permanent Memory System - Remembers project context across sessions (93% token savings)
• 20 Specialized Agents - From frontend to security, AI to payments
• 6 Tool Adapters - Works with OpenCode, Claude Code, Cursor, Aider, Copilot, and any other tool
• Existing Project Support - Works on new AND existing codebases
• Intelligent Orchestration - CEO that manages the entire lifecycle
• Parallel Execution - Multiple agents work simultaneously

Quick Start

Step 1: Copy Memory to Your Project

cp -r memory/ .memory/

Step 2: Choose Your Tool

Tool	Adapter	Instructions
OpenCode	adapters/opencode.md	Use SKILL.md
Claude Code	adapters/claude-code.md	Use CLAUDE.md
Cursor	adapters/cursor.md	Use .cursorrules
Aider	adapters/aider.md	Use .aider.conf.yml
GitHub Copilot	adapters/copilot.md	Use copilot-instructions.md
Any Tool	adapters/generic.md	Manual setup

Step 3: Use

You: "I want to create a SaaS for gym workouts"
AI: [Loads memory] -> Analyzes -> Plans -> Builds team -> Executes

How It Works

1. Memory System

The orchestrator maintains permanent memory in .memory/:

• project.md - Stack, structure, conventions
• errors.md - Mistakes to avoid
• successes.md - What works well
• progress.md - Current status
• decisions.md - Architecture decisions

Benefit: 93% fewer tokens per session, no re-analysis needed.

2. Agent Team (5-20 agents)

Core Team (always present)

• planner - Analysis and planning
• builder - Code execution
• reviewer - QA and code review
• documenter - Documentation

Specialized Teams (created as needed)

Team	Agents
Frontend	ui-specialist, frontend-specialist, mobile-specialist
Backend	backend-specialist, database-specialist, auth-specialist
Features	payment-specialist, ai-engineer, realtime-specialist, integration-specialist
Quality	testing-specialist, performance-specialist, security-specialist
Operations	devops-specialist, data-specialist

3. Workflow

1. Load Memory -> Understand context
2. Analyze Project -> Identify gaps
3. Create Plan -> Define milestones
4. Build Team -> Select specialists
5. Execute Tasks -> Work in parallel
6. Verify Quality -> Test and review
7. Update Memory -> Save progress
8. Deliver Results -> Report to user

Supported Tools

This system works with any coding tool. We provide adapters for:

Tool	Adapter File	How It Works
OpenCode	adapters/opencode.md	SKILL.md auto-loads orchestrator
Claude Code	adapters/claude-code.md	CLAUDE.md contains instructions
Cursor	adapters/cursor.md	.cursorrules loads rules
Aider	adapters/aider.md	.aider.conf.yml configures memory
GitHub Copilot	adapters/copilot.md	copilot-instructions.md loads rules
Other Tools	adapters/generic.md	Manual setup instructions

For Existing Projects

The system works on existing projects too:

1. Copy .memory/ to your project
2. The AI reads your codebase and fills memory automatically
3. It understands your stack, conventions, and patterns
4. It plans improvements and executes them

You: "Help me add payment feature to my existing app"
AI: [Reads your code] -> [Fills memory] -> [Plans] -> [Executes]

Supported Stacks

JavaScript/TypeScript

• Next.js 14+ (App Router)
• React + Vite
• Vue.js + Nuxt
• Node.js + Express
• tRPC

Python

• FastAPI
• Django
• Flask

Go

• Gin/Echo/Fiber

Rust

• Actix-web/Axum

Mobile

• React Native + Expo
• Flutter

Database

• PostgreSQL
• MySQL/MariaDB
• MongoDB
• Redis

Deploy

• Vercel
• Railway
• Fly.io
• AWS/GCP/Azure

Examples

Example 1: New SaaS

You: "I want to create a SaaS for project management"

AI creates:
- Stack: Next.js + PostgreSQL + Prisma
- Team: 10 specialists
- Plan: 6 weeks
- Files: 50+

Example 2: Existing Project

You: "Add payment feature to my app"

AI:
- Reads your existing codebase
- Identifies current stack (e.g., Next.js + Prisma)
- Creates payment-specialist
- Implements Stripe integration
- Updates memory with new decisions

Example 3: Continue Previous Work

You: "Continue with the next phase"

AI:
- Reads .memory/progress.md
- Knows exactly where you left off
- Resumes work from last task
- Updates memory with new progress

Project Structure

For OpenCode

your-project/
├── .opencode/
│   ├── agents/          # Project-specific agents
│   └── memory/          # Permanent memory
├── .memory/             # Universal memory (symlink or copy)
└── PLAN.md

For Claude Code

your-project/
├── .memory/             # Permanent memory
├── CLAUDE.md            # Instructions with memory loading
└── PLAN.md

For Cursor

your-project/
├── .memory/             # Permanent memory
├── .cursor/
│   └── agents/          # Project-specific agents
├── .cursorrules         # Instructions
└── PLAN.md

For Any Tool

your-project/
├── .memory/             # Permanent memory
├── .agents/             # Project-specific agents
└── PLAN.md

Token Efficiency

Scenario	Without Memory	With Memory
New session	5000+ tokens	500 tokens
Project analysis	3000 tokens	0 tokens
Error context	2000 tokens	200 tokens
Total	~10000 tokens	~700 tokens

Savings: ~93%

Documentation

• README.md - This file
• ADVANCED.md - Advanced usage
• docs/WORKFLOW.md - Detailed workflow
• docs/CUSTOMIZATION.md - Customization guide
• adapters/ - Tool-specific adapters

Contributing

1. Fork the repo
2. Create feature branch
3. Add new agent templates or adapters
4. Test with real projects
5. Submit PR

License

MIT License

Support

• Documentation: README.md
• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

Ready to build 4x faster? Start talking to the AI!