Agent Architecture — Multi-Agent Design
The Core Idea
One general-purpose AI agent can’t do everything well. Instead, I run a central intelligence (Legacy) supported by 8 domain-specific agents, each responsible for a quadrant or sub-domain of life.
Agent Roster
| Agent | Quadrant | Responsibility |
|---|---|---|
| Legacy | All | Central intelligence, orchestration, communication |
| Spartan | HP 27-28 | Body, fitness, physical mastery |
| Oracle | HP 25-26, 29-34 | Mind, money, meaning, personal growth |
| Eden | LE | Food growing, agriculture, sustainability |
| Empire | GE 75-84 | Business strategy, wealth building |
| Haven | LE 50-58 | Home, family, estate management |
| Guardian | UI 11 + HP 31 + LE 57 | Security, legal, protection |
| Grid | UI 00-12 | Infrastructure, DevOps, automation |
| Trinity | GE 75 + Community | Community building, public presence |
How They Communicate
All agents connect through a message bus (NATS JetStream). Legacy dispatches tasks, agents report results. Every agent can read the codex for context.
Why This Matters for Your Codex
The domain numbering system (00-99) maps directly to agent responsibilities. When you assign a domain to an agent, it knows exactly which folders to read, what context to load, and what kind of tasks it handles.