Tools & SetupFree Guide · April 2026

The Miss AI
Obsidian
Guide

How to build your AI second brain. Setup. Structure. Data. Connections. Claude Code.

Keira Nesdale · Miss AI

By Keira Nesdale Tools & Setup 30 min read
↓ Download PDF
01

What Is
Obsidian?

Obsidian is a note-taking app with a superpower.

Everything you write is stored locally on your computer in a folder called your Vault, just plain markdown files. No cloud lock-in. No subscription required. No one else owns your data. You can open every file in any text editor on earth and read them directly.

That sounds simple. Here is the part that changes things.

When you connect Obsidian to an AI agent like Claude Code, it becomes your second brain. The AI reads your notes, builds connections between them, writes new ones from your raw sources, and helps you think at a level that was never possible before.

Accessobsidian.md, Mac, Windows, Linux, iOS, Android
SetupDownload and create a vault. Under 5 minutes.
Free tierYes. Free for personal use, forever.
Paid plan$50 USD/year for sync across devices. Optional.
Best forFounders, creators, consultants, knowledge workers
The one-sentence version

Obsidian stores everything locally on your computer in a folder called your Vault, just plain markdown files, no cloud, no lock-in. Hook it up to an AI agent and it becomes your second brain.

02

Why You
Should Use It

Most people are losing their best thinking.

Ideas go into chat windows and disappear. Meeting notes live in ten different apps. Conversations with AI never compound, you start from scratch every single session. The insight you had three months ago is gone. The founder you met last year is a vague memory. The strategy you wrote in a Google Doc is buried somewhere you will never find it.

Obsidian fixes this. Here is why it is the right tool for anyone building something:

What makes it different from Notion

Notion is a great tool. But your data lives in Notion's cloud. If Notion changes pricing or closes, you lose access. Every note in Obsidian is a plain text file sitting on your hard drive. You own it completely. You can back it up, move it, or open it in any app. That independence is the foundation everything else is built on.

03

The Core Idea:
Karpathy's
LLM Wiki

In April 2026, Andrej Karpathy, co-founder of OpenAI and the person who coined vibe coding, went massively viral for sharing how he now uses LLMs to build personal knowledge wikis instead of just generating code.

Raw documents go into a folder. An AI reads them, extracts key insights, and writes structured interconnected notes. The knowledge compounds over time instead of being rediscovered from scratch on every question.

The problem with normal AI

When you use ChatGPT, Claude, or any AI tool with your documents, it works like RAG — Retrieval Augmented Generation. You upload files, the AI finds relevant chunks at query time, it generates an answer. Every time you ask a question it starts from zero. Nothing accumulates. Nothing compounds.

The wiki solution

Instead of retrieving from raw documents at query time, the AI incrementally builds and maintains a persistent wiki. When you add a new source, it reads it, extracts key information, and integrates it into the existing structure — updating pages, creating links, noting contradictions, strengthening the knowledge base.

The key insight

The wiki is a persistent, compounding artifact. The cross-references are already there. The contradictions have already been flagged. The synthesis already reflects everything you have read. Your knowledge gets richer with every source you add.

The three layers
LayerWhat it is
raw/Your curated collection of source documents. Articles, exports, transcripts, voice notes. Append-only. The AI reads from here but never edits.
wiki/LLM-generated markdown files. Summaries, entity pages, concept pages. The AI owns this layer. You read it.
CLAUDE.mdThe schema that tells the AI how your wiki is structured and what workflows to follow. You and the AI co-evolve this over time.
In practice

Obsidian is your IDE. Claude Code is your programmer. Your vault is the codebase.

04

Your Vault
Structure

This structure was built by combining two ideas from two people worth following.

Andrej Karpathy (@karpathy on X) published his LLM Wiki concept in April 2026, showing how AI agents can incrementally build and maintain a personal knowledge base from raw source material.

Remy (@aiwithremy on Instagram) shared his Obsidian business setup, showing how to structure a multi-brand founder vault with the _context pattern, Areas, and Projects. His system is built for real business operations, not just research.

The prompt I used

I am building an Obsidian vault and want to use it as my AI second brain. I run multiple brands, I have clients, I do coaching, and I create content. Here is how my business is structured: [describe your business]. Based on Karpathy's LLM Wiki concept and a business-first folder approach, can you build me a complete vault structure with folders, subfolders, and _context docs tailored to how I actually work?

Build your own vault structure

Copy and paste this into Claude to generate a structure tailored to your business:

I want to set up an Obsidian vault as my AI second brain.
Please build me a complete vault structure based on how
I actually work. Here is my situation:

My businesses / brands: [list them]
My income streams: [e.g. clients, products, coaching, content]
My content platforms: [e.g. Instagram, TikTok, YouTube, X]
My team size: [solo / small team / etc]
Tools I currently use: [Notion, Google Docs, Fireflies, etc]

Please create:
1. A top-level folder structure (6 folders max)
2. Subfolders inside Areas/ for each of my brands/businesses
3. A _context folder inside each Area with a list of docs
   I should create for that brand
4. A sample Brand Voice.md template I can fill in
5. A daily log template suited to how I work

Base the structure on Karpathy's LLM Wiki pattern
(raw/, wiki/, schema) combined with a business-first
folder approach where each brand has its own _context
folder as the single source of truth.
Top level folders
FolderWhat goes here
Areas/Your business areas and brands. Each gets its own folder with a _context subfolder.
Projects/One flat markdown file per project. Status: active, paused, or done.
Daily Logs/One note per day. Brain dump, tasks, wins, blockers, content ideas.
Meeting Notes/Notes from every call and meeting.
Notes/Evergreen reference notes. SOPs, research, guides. Includes a raw/ subfolder.
_System/Templates and attachments. Underscore pins it to the bottom of the sidebar.
Inside Areas/ — example for a multi-brand founder
Areas/
  Content Brand/    # Your personal brand and content
    _context/       # Brand voice, ICP, positioning, platforms
    Content/        # Scripts, captions, articles, reels
    Newsletter/     # Weekly issues, structure guide
    Analytics/      # Platform stats, what is working
  Product Business/ # Your SaaS, tool, or marketplace
    _context/       # Positioning, pricing, ICP
    Products/       # Product 1, Product 2, Product 3
  Clients/
    01-Pipeline/    # Leads with initial conversations
    02-Active/      # Signed clients in delivery
    03-Complete/    # Finished engagements
  Coaching/
    _context/       # Offer, pricing, target student
    Students/
  Partnerships/
    _context/       # Partnership context and overview
    Contacts/       # One note per key contact
Why the _context pattern matters

When something changes — your pricing, your ICP, your positioning — you update one file in _context. Every AI agent, every workflow, every template that reads from it gets the update automatically. No hunting through 20 folders to find the same document duplicated everywhere.

05

Step-by-Step
Setup

What you need before starting
Daily log template
# Daily Log — <% tp.date.now("YYYY.MM.DD") %>

## Brain dump
<!-- Just talk. What is on your mind today? -->

## What I am working on today
-

## Wins from yesterday
-

## Stuck on / blockers
-

## Follow ups needed
-

## Content ideas
-

## Random thoughts
<!-- Voice note style. No filter. -->
Important

Do not skip the _context docs. They are what makes your AI sound like you instead of a generic chatbot. Spend 20 minutes filling them in properly on day one. It pays back immediately on every piece of content you generate.

06

Plugins

Obsidian has hundreds of community plugins. Start with these and add more only when you have a specific reason.

PluginWhat it does
TemplaterMakes your templates smart. Auto-fills dates, creates notes from templates when you open a folder. Required for your daily log to auto-date.
Daily Notes (built-in)Creates a new daily note from your template every morning. Already in Obsidian, just needs to be turned on in Core Plugins.
CalendarAdds a calendar to your sidebar. Click any date to jump to that day's log.
Bases (built-in)Notion-style database views over your markdown files. Turns your Projects folder into a kanban or table view. Already in Obsidian.
How to install community plugins
1
Go to Settings in Obsidian (gear icon bottom left)
2
Click Community Plugins
3
Turn off Safe Mode
4
Click Browse
5
Search for the plugin name → Install → Enable
07

What Data
to Import

Your vault should not start empty. You have years of valuable thinking scattered across different platforms. Here is how to get it all in one place.

The rule

Everything goes into Notes/raw/ first. Do not organise it. Do not rename it. Just dump it in. Claude Code processes and sorts everything later. Your job is capture. The AI's job is organisation.

AI conversation history
PlatformHow to export
Claudeclaude.ai → Settings → Data and Privacy → Export Data. Emails you a zip of all your conversations.
ChatGPTchat.openai.com → Settings → Data Controls → Export Data. Emails you a zip.
Grokx.com → Settings → Privacy → Request Archive. Downloads as JSON.
GeminiIncluded in Google Takeout. Select Google Drive and Gemini when setting up your export.
Professional data
SourceHow to export
Google Docsdrive.google.com → search and filter by Docs → select all → right click Download. Comes as a zip of docx files.
Google Takeouttakeout.google.com → deselect all → select Drive, Meet, NotebookLM → Create export. Emails you a download link.
NotionSettings → Export all workspace content → Markdown format. Comes as zip of .md files that open perfectly in Obsidian.
LinkedInSettings → Data Privacy → Get a copy of your data → Request archive.
AirtableOpen each base → download as CSV from the grid menu.
Social media archives
PlatformHow to export
X / TwitterSettings → Your Account → Download an archive of your data.
InstagramAccounts Center → Your information and permissions → Export your information.
TikTokProfile → Settings → Privacy → Download your data.
YouTubeIncluded in Google Takeout. Select YouTube when setting up.
Meeting transcripts
08

Links and the
Graph View

Links are what turns Obsidian from a notes app into a second brain.

When you connect notes together, the Graph View shows you a visual map of how your ideas, projects, people, and businesses relate to each other. The more links you create, the more your vault looks like a living system rather than a folder of documents.

How to create a link

Type double square brackets around any word or phrase:

[[Your Product]] connects to [[Your Brand]] which feeds into [[Your Audience]]

Every bracketed term becomes a node on your graph. If that note does not exist yet, Obsidian creates a placeholder. When two notes both reference the same term, a line appears between them.

Add links to every _context doc

At the bottom of your Brand Voice doc add a links section like this:

[[Your Product]] [[Your Brand]] [[Podcast Name]] [[Newsletter]]
[[Coaching]] [[Platform 1]] [[Platform 2]] [[Client A]] [[Client B]]

Do the same for every _context doc and every project file. The more links you add, the richer your graph becomes.

Opening the Graph View

Press Cmd+G on Mac or Ctrl+G on Windows at any time. You will see every note as a dot and every link as a line. Notes with more connections appear larger and sit closer to the centre.

The compounding effect

The graph gets genuinely useful around 50 connected notes. At 200 notes it starts revealing unexpected connections between ideas you never consciously made. At 500 notes it becomes an asset you cannot replace. Start now.

09

Connecting
Claude Code

This is where everything comes together and you create the beautiful brain.

Claude Code is an AI agent that reads and writes files on your computer. When you point it at your Obsidian vault, it becomes the engine that processes your raw data, builds your wiki, creates links between notes, and maintains your second brain automatically.

What you need
How to install
npm install -g @anthropic-ai/claude-code

Navigate to your vault folder:

cd '/path/to/your/vault'

Launch Claude Code:

claude
The prompt to process your raw/ folder

Save this as a note in _System/Claude Code Prompts.md and paste it into Claude Code whenever you have new files to process:

There are new files in Notes/raw/ - please process all
unprocessed files, extract key insights, people, projects,
and ideas, and create properly linked markdown notes in
the correct folders:

- People and contacts  -> Areas/Partnerships/Contacts/
- Brand content        -> Areas/[Your Brand]/Content/
- Business insights    -> Notes/
- Client info          -> Areas/Clients/
- Podcast guests       -> Areas/Podcasts/

Use [[double bracket links]] to connect everything to
related notes already in the vault.
Update index.md and log.md if they exist.
The prompt to build connections

Once your vault has notes in it, run this to make your graph light up with connections:

Please read every note in my vault and add [[double bracket
links]] wherever related concepts, people, projects, or
brands are mentioned. Specifically:

- Link every mention of a project to its project file
- Link every mention of a person to their contact note
- Link every mention of a brand or product to its _context doc
- Link every mention of a platform to a platform note
- Link every mention of a client to their client folder

After adding links, update the graph by ensuring index.md
lists every note with a one-line summary.

The goal: every note should have at least 2 outbound links.
Start with the _context docs, then Projects, then Daily Logs.
How often to run it

You do not need to run Claude Code every time you add a file to raw/. Dump things in throughout the week — articles, exports, voice note transcripts, meeting notes — and run Claude Code once at the weekend to process everything in one pass. The graph expands, the links are created, your vault grows.

10

Your Daily
Workflow

Once your vault is set up, using it should feel effortless. Here is the rhythm that makes it work.

Every morning
1
Open Obsidian. Today's daily log auto-creates from your template.
2
Brain dump anything on your mind into the Brain Dump section.
3
Write your top 3 things to work on today.
4
Check your Follow Ups Needed section from yesterday.
Throughout the day
1
When you find a useful article, click Web Clipper to save it to Notes/raw/.
2
After any meeting, paste your notes into Meeting Notes/.
3
When you have a content idea, drop it in today's daily log.
4
If you record a voice note, paste the transcript into your daily log.
Weekly
1
Run Claude Code over your raw/ folder to process new content.
2
Review your Projects/ folder and update statuses.
3
Open the graph view and see what new connections have formed.
4
Follow up on anything sitting in your 01-Pipeline client folders.
The 10 mistakes most people make
MistakeWhat to do instead
Spending hours organising raw/Never organise raw/. Dump everything in and let Claude Code sort it.
Building perfect structure before adding notesStart with 6 folders and your first _context doc. Structure grows with content.
Not adding links between notesLinks are the whole point. Add them to every _context doc and project file from day one.
Importing everything including old junkOnly import what you would actually go looking for. Quality over volume.
Using Obsidian as a basic notes appConnect Claude Code. The vault alone is useful. With Claude Code it is a different category of tool.
Skipping the _context docsThey are your AI's briefing. Without them every response is generic. With them it sounds like you.
Never opening the graph viewOpen it weekly. It shows connections you did not know existed.
Waiting until it is perfect to startOpen Obsidian today. Create 6 folders. Add one note. Done. Everything else is iteration.
Keep building

The graph you see today with a handful of connected dots? In six months it will look like a galaxy. Every note you add, every link you create, every export you import makes it smarter and more useful.

The people who get the most from Obsidian are not the most organised. They are the most consistent. They add one note a day, run Claude Code once a week, and follow up on what the vault surfaces. That compounding effect is real.

Start simple. Stay consistent. Let the AI do the heavy lifting.

K
Keira Nesdale
@RealMissAI · Miss AI Podcast
← Back to vault