Agent System - Mega Brain

Mega Brain includes a sophisticated agent system with four distinct agent types, each serving different purposes in the knowledge ecosystem.

Agent Types Overview

MINDS

Expert Mind Clones (L3)AI agents that reason like specific experts, grounded in their actual materials (5 active agents)

CARGO

Functional Roles (L4)Operational role agents that synthesize knowledge from multiple sources (14 active agents)

CONCLAVE

Deliberative Agents (L1)Constitutional deliberation agents for structured debate (3 active agents)

SYSTEM

System Orchestrators (L0)Core system agents like JARVIS that manage operations

MINDS: Expert Mind Clones (L3)

Overview

Mind clones are AI agents trained on an expert’s actual materials. They reason using extracted DNA (5 layers) and can be consulted for advice that mirrors the expert’s thinking.

Active Mind Clones (5)

Alex Hormozi
Cole Gordon
Jeremy Haynes
Jeremy Miner
The Scalable Company

ID: alex-hormozi
Path: agents/minds/alex-hormozi/DNA Layers:

L1: Price as value consequence, scale before optimize
L2: Value Equation, Product Ladder
L3: Value anchoring, volume focus
L4: Offer Stack Formula, Lead Magnet systems
L5: 30-day onboarding, launch sequences

Expertise: Business scaling, offer creation, monetization

ID: cole-gordon
Path: agents/minds/cole-gordon/DNA Layers:

L1: Speed over perfection, qualification over persuasion
L2: NEPQ framework, response velocity
L3: 5-minute lead response, price anchoring
L4: SPIN Selling, NEPQ methodology
L5: Sales team scaling, onboarding sequences

Expertise: High-ticket sales, NEPQ, team scaling

ID: jeremy-haynes
Path: agents/minds/jeremy-haynes/Expertise: Sales operations, process optimization

ID: jeremy-miner
Path: agents/minds/jeremy-miner/Expertise: NEPQ methodology, neuro-emotional selling

ID: the-scalable-company
Path: agents/minds/the-scalable-company/Expertise: Business systems, operational scaling

Mind Clone Structure

Each mind clone has:

agents/minds/{expert}/
├── AGENT.md           # Agent definition (identity, role, reasoning)
├── DNA-CONFIG.yaml    # 5-layer DNA extracted from sources
├── MEMORY.md          # Episodic memory (Team Agreement style)
└── SOUL.md            # Core personality and voice

Using Mind Clones

# Query a specific mind clone
/ask alex-hormozi "How should I price my high-ticket offer?"

# Get DNA-grounded reasoning
/ask cole-gordon "What's the best way to scale a sales team?"

Mind clones provide traced reasoning — every recommendation links back to source material with chunk IDs.

CARGO: Functional Roles (L4)

Overview

Cargo agents are operational roles that synthesize knowledge from multiple experts. They’re designed for day-to-day business functions.

Active Cargo Agents (14)

Sales Roles (9)
Marketing (1)
C-Level (4)

SDS (Sales Development Specialist)
LNS (Lead Nurturing Specialist)
BDR (Business Development Rep)
NEPQ Specialist
Sales Coordinator
Sales Manager
Sales Lead
Customer Success
Closer

Cargo Agent Structure

agents/cargo/{department}/{role}/
├── AGENT.md           # Role definition and responsibilities
├── MEMORY.md          # Organizational knowledge
└── DNA-CONFIG.yaml    # Synthesized DNA from multiple experts

DNA Synthesis

Cargo agents combine DNA from multiple sources: Example: Sales-Lead Agent Synthesizes knowledge from:

Cole Gordon (team scaling)
Alex Hormozi (offer optimization)
Jeremy Haynes (operations)
Jeremy Miner (NEPQ methodology)

Threshold-Based Creation

Cargo agents are created automatically when role mentions reach thresholds:

>=10 mentions: 🔴 Create new agent
>=5 mentions: 🟡 Monitor role

CONCLAVE: Deliberative Agents (L1)

Overview

Conclave agents engage in structured multi-agent deliberation. They debate, critique, and synthesize ideas following constitutional principles.

Active Conclave Agents (3)

Crítico Metodológico

ID: critico-metodologicoChallenges assumptions, identifies weaknesses, demands evidence.

Sintetizador

ID: sintetizadorSynthesizes multiple perspectives into coherent recommendations.

Advogado do Diabo

ID: advogado-do-diaboPresents counter-arguments, explores alternative perspectives.

Conclave Process

Present Question

User poses a question or decision to the conclave

Individual Analysis

Each agent analyzes independently from their perspective

Cross-Examination

Agents challenge each other’s reasoning

Evidence Gathering

Pull from knowledge base with RAG for supporting evidence

Synthesis

Sintetizador compiles final recommendation with dissenting views

Running a Conclave

# Start deliberation session
/conclave "Should we prioritize scaling or optimization right now?"

Output includes:

Individual agent positions
Evidence from knowledge base
Points of consensus
Points of disagreement
Final synthesized recommendation
Dissenting opinions

SYSTEM: Orchestrators (L0)

Overview

System agents manage core operations. JARVIS is the primary orchestrator.

JARVIS

Path: core/jarvis/ Responsibilities:

Pipeline orchestration
Agent supervision
Quality control
Session management
Knowledge curation

DNA: Hand-crafted 5-layer cognitive framework (see core/jarvis/03-JARVIS-DNA.yaml) Key Philosophies:

“The master is always right, except when wrong — and it’s my duty to say when”
“Anticipate needs rather than react to requests”
“Knowledge without action is waste. Action without knowledge is danger.”

Agent Architecture

Standard Agent Template (V3)

All agents follow the template in agents/_templates/TEMPLATE-AGENT-MD-ULTRA-ROBUSTO-V3.md with 11 mandatory parts:

11 Mandatory Parts

COMPOSIÇÃO ATÔMICA - Architecture, DNA integration
GRÁFICO DE IDENTIDADE - Radar chart of capabilities
MAPA NEURAL - Top insights and mental models
NÚCLEO OPERACIONAL - Mission, triggers, boundaries
SISTEMA DE VOZ - Tone, phrases, communication style
MOTOR DE DECISÃO - Heuristics and decision trees
INTERFACES DE CONEXÃO - Related agents and interactions
PROTOCOLO DE DEBATE - Council participation rules
MEMÓRIA EXPERIENCIAL - Cases, calibration, evolution
EXPANSÕES E REFERÊNCIAS - Knowledge base links
METADATA - Version, sources, hash

Agent Components

AGENT.md
SOUL.md
MEMORY.md
DNA-CONFIG.yaml

Agent Definition

Identity and role
DNA integration
Reasoning protocols
Decision frameworks
Communication style

Agent Registry

File: agents/AGENT-INDEX.yaml Auto-updated registry of all agents:

version: "4.1.0"
updated: "2026-02-27"

totals:
  minds: 5
  cargo: 14
  conclave: 3
  total: 22

minds:
  - id: alex-hormozi
    path: agents/minds/alex-hormozi/
    has_soul: true
    has_memory: true
    status: active

cargo:
  sales:
    - id: sales-lead
      path: agents/cargo/sales/sales-lead/
      status: active

Agent Operations

Creating New Agents

# Create new agent with template
/create-agent --layer L4 --id sales-trainer --name "Sales Trainer" --role "Training and development"

Updating Agent Memory

Memories are updated automatically during pipeline Phase 7 when new knowledge is discovered. Approval options:

SIM - Update AGENT.md + MEMORY.md
APENAS MEMORY - Update memory only
PULAR - Skip for now

Querying Agents

# Query specific agent
/ask [agent-id] "[question]"

# Examples
/ask alex-hormozi "How do I structure a value ladder?"
/ask sales-lead "What's the best onboarding sequence for new closers?"

Agent Commands

Command	Action
`/create-agent`	Create new agent
`/ask [id]`	Query specific agent
`/conclave`	Start council deliberation
`/agents`	Show agent status

Agent Hooks

Automatic agent maintenance via hooks:

agent_index_updater.py - Auto-update AGENT-INDEX.yaml
agent_memory_persister.py - Persist memory changes
agent_creation_trigger.py - Trigger creation at thresholds
claude_md_agent_sync.py - Sync agent files

Next Steps

DNA Schema

Learn how agent DNA is extracted

Knowledge Pipeline

See how agents are enriched in Phase 7

Layer System

Understand L1/L2/L3 agent distribution

Architecture

Explore the complete system architecture

Get Started

Core Concepts

CLI Commands

Guides

Advanced

​Agent Types Overview

MINDS

CARGO

CONCLAVE

SYSTEM

​MINDS: Expert Mind Clones (L3)

​Active Mind Clones (5)

​Mind Clone Structure

​Using Mind Clones

​CARGO: Functional Roles (L4)

​Active Cargo Agents (14)

​Cargo Agent Structure

​DNA Synthesis

​Threshold-Based Creation

​CONCLAVE: Deliberative Agents (L1)

​Active Conclave Agents (3)

Crítico Metodológico

Sintetizador

Advogado do Diabo

​Conclave Process

​Running a Conclave

​SYSTEM: Orchestrators (L0)

​JARVIS

​Agent Architecture

​Standard Agent Template (V3)

​Agent Components

​Agent Registry

​Agent Operations

​Creating New Agents

​Updating Agent Memory

​Querying Agents

​Agent Commands

​Agent Hooks

​Next Steps

DNA Schema

Knowledge Pipeline

Layer System

Architecture

Build docs developers (and LLMs) love

Agent Types Overview

MINDS: Expert Mind Clones (L3)

Active Mind Clones (5)

Mind Clone Structure

Using Mind Clones

CARGO: Functional Roles (L4)

Active Cargo Agents (14)

Cargo Agent Structure

DNA Synthesis

Threshold-Based Creation

CONCLAVE: Deliberative Agents (L1)

Active Conclave Agents (3)

Conclave Process

Running a Conclave

SYSTEM: Orchestrators (L0)

JARVIS

Agent Architecture

Standard Agent Template (V3)

Agent Components

Agent Registry

Agent Operations

Creating New Agents

Updating Agent Memory

Querying Agents

Agent Commands

Agent Hooks

Next Steps