Architecture
Key Features
Multi-Agent System
7 specialized agents orchestrated by supervisor
Sequential Pipeline
PRD → Data → Design → Logic → API → Frontend → Deployment
Human-in-the-Loop
Strategic interrupt points for review and approval
Artifact Generation
Complete specifications and code artifacts
Graph Execution
Parallel and sequential agent execution
Streaming Responses
Real-time progress updates via SSE
Agent System
The system consists of 7 specialized agents plus a supervisor:- PRD Agent - Product Requirements Document generation
- Data Agent - Data modeling and schema design
- Design Agent - UI/UX design system and tokens
- Logic Agent - Business logic and workflows
- API Agent - REST/GraphQL API specifications
- Frontend Agent - React component architecture
- Deployment Agent - Infrastructure and CI/CD pipelines
Execution Phases
The system executes through sequential phases:- PRD - Product requirements and user stories
- Data - Entity models and relationships
- Design - Design system and UI tokens
- Logic - Business rules and workflows
- API - API endpoints and schemas
- Frontend - Component hierarchy and state management
- Deployment - Infrastructure and deployment configs
Human-in-the-Loop (HITL)
Strategic interrupt points allow human review:- After PRD - Review requirements before data modeling
- After Logic - Review business rules before API design
- After API - Review API spec before frontend implementation
Quick Start
Prerequisites
- Bun (recommended) or Node.js
- Cloudflare Account with Workers access
- Gemini API Key (for agent execution)
Installation
Environment Setup
Create.env.local:
Development
Project Structure
Related Documentation
- Supervisor API - Supervisor worker API
- Agent API - Agent workers API
- Frontend Canvas - Agent canvas interface
- Backend Supervisor - Supervisor implementation
- Backend Agents - Agent implementations
- Design Architecture - System architecture