The Mini AGI Blueprint

The Promise

What if your AI assistant actually knew you?

Not the surface-level "you mentioned you like coffee" kind of knowing. I mean really knew you:

• Every project you're working on
• Your goals for the next year
• Your writing voice and preferences
• The ideas you captured three months ago but haven't developed yet
• The patterns in how you think and work

This guide shows you how to build exactly that.

Why This Matters

The Problem Everyone Has

AI assistants are stateless. They forget everything between sessions.

You explain your project. You describe your goals. You share your preferences. Then the next day, you start over from zero.

This creates two massive problems:

The Solution

Give AI access to a persistent knowledge base that YOU control.

Instead of hoping AI will somehow retain context, give it access to a structured system that you maintain. Your accumulated knowledge, your projects, your preferences - all in one place that AI can reference.

The key insight: AI doesn't need to "remember" if it can "look up."

What This System Does

When I open a terminal and start Claude Code, it has access to:

• My entire Second Brain - 500+ notes across projects, ideas, philosophies, daily captures
• Project files and context - What I'm building, the status of each project, my todo lists
• Brand strategy and voice guides - How I want to communicate, what I stand for
• Historical context - Past decisions, learnings, patterns I've identified
• Daily captures - Raw thoughts, ideas, observations from throughout each day

The Architecture

My setup uses three components working together:

Step-by-Step Setup Guide

Second-Brain/
├── Daily Notes/
│   └── 2026-01-22.md
├── Projects/
│   └── My-Project/
│       ├── _PROJECT_PLAN.md
│       └── _TODOS.md
├── Knowledge/
│   └── (your learnings and reference material)
└── CLAUDE.md

{
  "mcpServers": {
    "second-brain": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/path/to/your/Second-Brain"
      ]
    }
  }
}

# CLAUDE.md - System Instructions

## Who I Am

[Your name and what you're building. 2-3 sentences.]

## How This System Is Organized

### Daily Notes
- Location: `Daily Notes/YYYY-MM-DD.md`
- Purpose: Raw captures, brain dumps, meeting notes

### Projects
- Location: `Projects/[Project-Name]/`
- Each project has:
  - `_PROJECT_PLAN.md` - Goals, vision, context
  - `_TODOS.md` - Active tasks

## How to Help Me

When I ask for help with content:
1. Check my content schedule if one exists
2. Reference my voice/style guides if available
3. Pull from my idea backlog for inspiration

## My Preferences

- I prefer [direct/conversational/formal] communication
- Always [preference 1]
- Never [anti-preference 1]

cd /path/to/your/Second-Brain
claude

Daily Workflows: How I Actually Use This

The Results

The biggest shift is psychological. I no longer feel like I'm working with an AI. I feel like I'm working with a system that knows me.

The compounding effect is real. Every note I write makes the system smarter. Every idea I capture is available later. Nothing falls through the cracks.

What's Next

Your Turn

You now have everything you need to build your own mini AGI:

• Obsidian or a markdown folder - Your knowledge base
• Claude Code - The AI interface
• Filesystem MCP - The connection layer
• CLAUDE.md - Your system prompt

The goal isn't a perfect system. It's a system that grows with you.