> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/thiagofinch/mega-brain/llms.txt
> Use this file to discover all available pages before exploring further.

# Working with Agents

> Query and interact with AI agents enriched with expert knowledge from your materials

Mega Brain creates AI agents that are automatically enriched with knowledge as you process materials. This guide shows you how to work with different agent types and get expert insights.

## Agent Types

Mega Brain has 4 categories of agents:

<CardGroup cols={2}>
  <Card title="Cargo Agents" icon="briefcase">
    Functional role agents (Sales, Marketing, Operations, Finance)
    **Count:** 29 active agents
  </Card>

  <Card title="Mind Clones" icon="brain">
    Expert agents that reason like specific people
    **Count:** 5 person agents
  </Card>

  <Card title="Conclave" icon="users">
    Multi-perspective deliberation council
    **Count:** 3 council agents
  </Card>

  <Card title="System Agents" icon="gear">
    JARVIS and internal system agents
    **Count:** 2 system agents
  </Card>
</CardGroup>

<Note>
  **Total Active Agents:** 22 across all categories
</Note>

## Viewing Agent Status

### List All Agents

```bash theme={null}
/agents
```

<Accordion title="Sample Output">
  ```
  ═══════════════════════════════════════════════
                AGENTS STATUS
                  2026-03-06
  ═══════════════════════════════════════════════

  🤖 ACTIVE AGENTS: 37

  C-LEVEL:
     CRO      │ MEMORY updated: 2026-03-04 │ Knows: 3 experts
     CFO      │ MEMORY updated: 2026-03-04 │ Knows: 2 experts
     CMO      │ MEMORY updated: 2026-03-01 │ Knows: 1 expert
     COO      │ MEMORY updated: 2026-03-04 │ Knows: 2 experts

  SALES:
     CLOSER           │ MEMORY: 2026-03-05 │ 3 experts │ v2.1
     BDR              │ MEMORY: 2026-03-05 │ 2 experts │ v1.8
     SDS              │ MEMORY: 2026-03-05 │ 2 experts │ v1.5
     LNS              │ MEMORY: 2026-03-04 │ 2 experts │ v1.3
     SALES-MANAGER    │ MEMORY: 2026-03-05 │ 3 experts │ v2.0
     SALES-LEAD       │ MEMORY: 2026-03-04 │ 2 experts │ v1.4
     SALES-COORDINATOR│ MEMORY: 2026-03-03 │ 1 expert  │ v1.0
     CUSTOMER-SUCCESS │ MEMORY: 2026-03-02 │ 1 expert  │ v1.1

  MIND CLONES:
     Alex Hormozi     │ DNA: 19 items │ Weight: 0.74
     Cole Gordon      │ DNA: 23 items │ Weight: 0.78
     Sam Ovens        │ DNA: 15 items │ Weight: 0.71

  ───────────────────────────────────────────────

  ⚠️ OUTDATED (>3 days):
     CMO, SALES-COORDINATOR, CUSTOMER-SUCCESS

  🚨 SUGGESTED FOR CREATION:
     🔴 APPOINTMENT-SETTER (12 mentions - threshold reached)
     🟪 COPYWRITER (8 mentions - monitor)

  ═══════════════════════════════════════════════
  ```
</Accordion>

### Filter by Category

```bash theme={null}
# View only sales agents
/agents --category SALES

# View only C-level
/agents --category C-LEVEL

# View outdated agents
/agents --outdated
```

### Check Expert Knowledge

```bash theme={null}
# Which agents know about Cole Gordon?
/agents --knowledge "Cole Gordon"
```

**Output:**

```
📚 AGENTS THAT KNOW Cole Gordon:

CLOSER
├─ Sources: CG001, CG002, CG003
├─ Insights: 23 (HIGH: 8, MEDIUM: 10, LOW: 5)
└─ Frameworks: NEPQ, Objection Loop, 7 Universal Closes

CRO
├─ Sources: CG001, CG003
├─ Insights: 12
└─ Frameworks: Farm System, Christmas Tree Structure

SALES-MANAGER
├─ Sources: CG001, CG002, CG003
├─ Insights: 18
└─ Frameworks: 28 Rules of Closing, Farm System

📄 DOSSIER AVAILABLE:
   knowledge/dossiers/persons/DOSSIER-COLE-GORDON.md
```

## Querying Cargo Agents

### Sales Agents

<Tabs>
  <Tab title="CLOSER">
    **Expertise:** High-ticket closing, objection handling, discovery calls

    ```bash theme={null}
    /ask closer "How do I handle the 'I need to think about it' objection?"
    ```

    **Response format:**

    ```
    ────────────────────────────────────
             CLOSER RESPONDS
    ────────────────────────────────────

    "I need to think about it" is Layer 3 of the
    Onion of Blame (Cole Gordon framework).

    It's NOT the real objection - it's a smoke screen.

    USE THE PEEL-BACK:

    "I understand you want to think about it. And
    I respect that. But just so I'm clear - when
    people say they need to think about it, it's
    usually because ONE of three things:

    1. They're not 100% sure this will work for them
    2. The timing doesn't feel quite right
    3. The investment feels like a stretch

    Which one of those is closest for you?"

    Then SHUT UP. Let them tell you the real objection.

    SOURCES:
    [chunk_CG001_067] Onion of Blame framework
    [chunk_CG002_034] Layer 3 = thinking objection
    [chunk_CG001_089] Peel-back technique

    CONFIDENCE: 0.87 (High)

    RELATED:
    - NEPQ framework for discovery
    - 7 Universal Closes
    ```
  </Tab>

  <Tab title="BDR (Business Development Rep)">
    **Expertise:** Cold outreach, lead qualification, appointment setting

    ```bash theme={null}
    /ask bdr "What's a good cold call opening?"
    ```

    **Response:**

    ```
    Pattern Interrupt → Permission → Reason for Call

    "Hey [Name], this is [Your Name] with [Company].
    I know I'm calling out of the blue - did I catch
    you at a bad time?"

    [Wait for response]

    "Great. The reason I'm calling is I work with
    [similar companies] who struggle with [specific problem].
    I'm not sure if that's relevant to you, but if it
    is, I'd love to ask you 2 quick questions to see
    if it makes sense to chat further. Fair?"

    KEY PRINCIPLES:
    1. Don't pretend you know them
    2. Ask permission ("bad time?")
    3. Use social proof ("companies like yours")
    4. Name the problem, not your solution
    5. Low commitment ask ("2 questions")

    SOURCES:
    [chunk_CG001_023] Cold call framework
    [chunk_CG001_024] Permission-based approach

    CONFIDENCE: 0.82
    ```
  </Tab>

  <Tab title="SALES-MANAGER">
    **Expertise:** Team structure, metrics, hiring, coaching

    ```bash theme={null}
    /ask sales-manager "How do I know if I need more closers or better closers?"
    ```

    **Response:**

    ```
    Use Cole Gordon's diagnostic:

    IF close rate < 60%:
      → Problem: SKILL (need better closers/training)
      → Don't hire more yet

    IF close rate >= 60% AND pipeline full:
      → Problem: CAPACITY (need more closers)
      → Safe to scale

    IF close rate >= 60% BUT pipeline not full:
      → Problem: LEAD GENERATION (need more leads)
      → Don't hire closers, fix top of funnel

    CURRENT BENCHMARKS:
    • Elite closers: 70%+ close rate
    • Good closers: 60-70%
    • Need coaching: 50-60%
    • Replace: <50%

    SOURCES:
    [chunk_CG001_045] 60% benchmark
    [chunk_CG002_078] Diagnostic framework

    CONFIDENCE: 0.91 (Very High)
    ```
  </Tab>
</Tabs>

### C-Level Agents

<Steps>
  <Step title="CRO (Chief Revenue Officer)">
    **Focus:** Revenue growth, sales strategy, pricing

    ```bash theme={null}
    /ask cro "Should I increase prices or volume?"
    ```

    CRO will analyze using:

    * Unit economics
    * Market positioning frameworks
    * Sales capacity constraints
    * Evidence from processed materials
  </Step>

  <Step title="CFO (Chief Financial Officer)">
    **Focus:** Financial metrics, profitability, risk

    ```bash theme={null}
    /ask cfo "Is my LTV:CAC ratio healthy?"
    ```

    CFO will evaluate:

    * Industry benchmarks (3:1 threshold)
    * Cash flow implications
    * Payback period
    * Sources: Hormozi financial frameworks
  </Step>

  <Step title="CMO (Chief Marketing Officer)">
    **Focus:** Positioning, messaging, demand gen

    ```bash theme={null}
    /ask cmo "How do I position against competitors?"
    ```

    CMO will suggest:

    * Differentiation strategies
    * Messaging frameworks
    * Channel recommendations
  </Step>

  <Step title="COO (Chief Operations Officer)">
    **Focus:** Systems, processes, scalability

    ```bash theme={null}
    /ask coo "How do I scale operations without breaking?"
    ```

    COO will address:

    * Process documentation
    * Bottleneck identification
    * Automation opportunities
  </Step>
</Steps>

## Querying Mind Clones

### Person-Specific Agents

Ask questions as if speaking directly to the expert:

<CodeGroup>
  ```bash Alex Hormozi theme={null}
  /ask alex-hormozi "How should I think about pricing?"

  # Responds using:
  # - Value Equation framework
  # - Grand Slam Offer principles  
  # - His specific philosophies
  # - Direct quotes from materials
  ```

  ```bash Cole Gordon theme={null}
  /ask cole-gordon "How do I improve my close rate?"

  # Responds using:
  # - NEPQ framework
  # - 28 Rules of Closing
  # - Specific techniques and scripts
  # - His coaching methodology
  ```

  ```bash Sam Ovens theme={null}
  /ask sam-ovens "How do I structure my consulting offer?"

  # Responds using:
  # - Setterlun frameworks
  # - Consulting delivery models
  # - Pricing strategies
  # - Client acquisition methods
  ```
</CodeGroup>

### Response Structure

Every mind clone response includes:

1. **Direct Answer** - In the expert's voice
2. **Framework Application** - Named methodologies used
3. **Sources** - Chunk\_ids linking to original material
4. **Confidence Score** - 0.00-1.00 based on evidence weight
5. **Related Concepts** - Additional relevant frameworks

<Tip>
  Mind clones only use knowledge with weight ≥ 0.70 to ensure high-confidence responses.
</Tip>

## Agent Auto-Enrichment

Agents are automatically updated as you process materials:

<Accordion title="How Auto-Enrichment Works">
  ### Phase 7: Agent Enrichment (Pipeline)

  When processing completes:

  ```
  1. Extract themes and frameworks from insights
  2. Map themes to relevant agents using THEME_TO_AGENTS:

     "02-PROCESSO-VENDAS" → [closer, SDS, LNS]
     "04-COMISSIONAMENTO" → [sales-manager, CRO, CFO]
     "07-PRICING" → [CRO, CFO, closer]

  3. For each mapped agent:
     - Load existing MEMORY-{agent}.md
     - Append new frameworks
     - Add source reference
     - Update last_modified timestamp

  4. Verify coverage with Agent Coverage Check
  ```

  ### Framework-Specific Routing

  Certain frameworks trigger specific agents:

  ```yaml theme={null}
  "3 Audience Buckets": [closer, SDS, LNS]
  "NEPQ Framework": [closer, SDS]
  "Email Hammer": [LNS]
  "Farm System": [sales-manager, CRO]
  "Christmas Tree Structure": [sales-manager, CRO]
  ```

  ### Coverage Verification

  Pipeline Phase 8.6 verifies:

  ```
  ☐ AGENT COVERAGE CHECK (Mandatory):

  1. Identify all themes in SOURCE_ID
  2. For each theme, lookup THEME_TO_AGENTS
  3. For each framework, lookup FRAMEWORK_TO_AGENTS
  4. List ALL agents that SHOULD receive content
  5. Verify grep -l "$SOURCE_ID" in each MEMORY-{agent}.md
  6. IF any agent missing SOURCE_ID:
     → LOG ERROR: "Agent Coverage Failed"
     → List expected vs. actual agents
     → EXIT with status: AGENT_COVERAGE_FAILED
  ```
</Accordion>

## Agent Creation Workflow

### Automatic Creation (Threshold-Based)

Agents are auto-created when roles reach thresholds:

<Steps>
  <Step title="Role Detection">
    As materials are processed, mentions of roles are counted:

    ```
    CLOSER: 21.6 weighted mentions → TIER 1 (Established)
    APPOINTMENT-SETTER: 12.3 mentions → TIER 1 (Create agent)
    COPYWRITER: 7.8 mentions → TIER 2 (Emerging, monitor)
    PODCAST-HOST: 2.1 mentions → TIER 3 (Emergent, observe)
    ```
  </Step>

  <Step title="Threshold Check">
    **TIER 1 (Established):** weighted\_score ≥ 10, sources ≥ 2
    → **Action:** CREATE AGENT

    **TIER 2 (Emerging):** weighted\_score ≥ 5, sources ≥ 1
    → **Action:** Monitor growth

    **TIER 3 (Emergent):** weighted\_score ≥ 3, sources ≥ 1\
    → **Action:** Observe
  </Step>

  <Step title="Auto-Creation Trigger">
    When TIER 1 reached:

    ```
    python scripts/agent_creator.py --check
    ```

    Creates:

    * `agents/cargo/SALES/AGENT-APPOINTMENT-SETTER.md`
    * `agents/cargo/SALES/MEMORY-APPOINTMENT-SETTER.md`
    * Corresponding SUA-EMPRESA files (org roles)
  </Step>
</Steps>

### Manual Creation

Create agents manually:

```bash theme={null}
/create-agent APPOINTMENT-SETTER --category SALES --sub-of SALES-MANAGER
```

<Warning>
  Manual creation bypasses threshold checks. Use for special cases only.
</Warning>

## Best Practices

<CardGroup cols={2}>
  <Card title="Ask Specific Questions" icon="bullseye">
    Better: "How do I handle the price objection?"
    Worse: "Tell me about sales"
  </Card>

  <Card title="Check Confidence Scores" icon="gauge-high">
    ≥ 0.80: High confidence, actionable
    0.70-0.79: Good confidence
    \< 0.70: Agent won't use (filtered out)
  </Card>

  <Card title="Review Sources" icon="book">
    Always check chunk\_ids to verify claims
    Trace back to original material context
  </Card>

  <Card title="Use Right Agent Type" icon="sitemap">
    Tactical questions → Cargo agents
    Expert perspective → Mind clones
    Strategic decisions → Conclave
  </Card>
</CardGroup>

## Troubleshooting

<Accordion title="Common Issues">
  **Issue: Agent says "I don't have knowledge about that"**

  Solutions:

  1. Check if agent has been enriched with relevant sources:
     ```bash theme={null}
     /agents --knowledge "[expert-name]"
     ```
  2. Process more materials on that topic
  3. Use a different agent (e.g., mind clone vs. cargo)

  **Issue: Response lacks sources/chunk\_ids**

  Causes:

  * Agent memory not properly updated
  * Pipeline didn't complete Phase 7
  * Sources missing from MEMORY-{agent}.md

  Fix:

  ```bash theme={null}
  # Reprocess materials to update agent memories
  /process-jarvis inbox/[person]/[file].txt
  ```

  **Issue: Agent is outdated (>3 days)**

  ```bash theme={null}
  # Check which agents need updates
  /agents --outdated

  # Process new materials to refresh
  /process-jarvis [new-materials]
  ```

  **Issue: Want to create agent but below threshold**

  ```bash theme={null}
  # Check current mentions
  /agents

  # Either:
  # 1. Process more materials mentioning that role
  # 2. Force creation (not recommended):
  /create-agent [ROLE] --force
  ```
</Accordion>

## Next Steps

<CardGroup cols={2}>
  <Card title="Run Conclave" icon="users" href="/guides/using-conclave">
    Multi-agent deliberation for strategic decisions
  </Card>

  <Card title="Create Mind Clones" icon="dna" href="/guides/creating-mind-clones">
    Extract expert DNA and create person-specific agents
  </Card>

  <Card title="Manage Sessions" icon="floppy-disk" href="/guides/session-management">
    Save and resume work sessions
  </Card>

  <Card title="Process Materials" icon="gears" href="/guides/processing-pipeline">
    Run the pipeline to enrich agent knowledge
  </Card>
</CardGroup>