Core Concepts
Core Concepts
Understanding these core concepts will help you use Shai effectively.
Overview
Shai is built around five key concepts:
- Cellular Development - Constraining agents to specific components
- Resource Sets - Defining collections of allowed resources
- Apply Rules - Mapping workspace paths to resource sets
- Selective Elevation - Controlled host-side operations
- How Shai Works - Understanding the architecture
Quick Mental Model
Think of Shai as establishing guardrails for your AI agents:
- Cellular Development: Define which areas agents can modify (they can read all workspace code, but only change specific parts)
- Resource Sets: Define what’s available in each area (network, credentials, host mounts)
- Apply Rules: Decide which areas get which resources
- Selective Elevation: Provide a phone to call outside when needed
- Architecture: How Shai enforces all these rules