TUTORIAL: OBSIDIAN + CLAUDE CODE SECOND BRAIN
You take notes. Lots of them. Meeting notes, ideas, research, project plans, random thoughts at 2 AM. They pile up in your vault, and then... nothing. They just sit there. You forget what you wrote. You make the same decision twice because you can't find the note from six months ago. Sound familiar?
What if your Obsidian vault could actually think with you? Not just store your notes, but connect ideas, surface forgotten context, generate daily plans, and get smarter over time?
That's exactly what happens when you pair Obsidian with Claude Code. This setup has been blowing up across social media, and for good reason: it solves the biggest problem with AI today - the need to constantly re-explain yourself every time you start a new chat.
This guide walks you through every step to build your own AI second brain from scratch.
Why Use Obsidian with Claude Code for a Second Brain?
Before we get into the setup, let's understand why this specific combination works so well.
Obsidian stores everything as plain Markdown files on your local machine. No proprietary formats, no cloud lock-in, no databases. Just folders of .md files sitting on your computer. This matters because Claude Code can read, write, and search plain text files directly - no special API or integration required.
Claude Code is an agentic AI tool that runs in your terminal. Unlike a chatbot that forgets everything after each conversation, Claude Code operates directly on your file system. Point it at your vault, and it instantly has access to everything you've ever written.
Together, they create a feedback loop: you take notes, Claude reads them, Claude generates new insights and connections, those become new notes, and the system gets smarter over time. Your vault stops being a graveyard and starts being a living knowledge base.
What You Need to Build an Obsidian AI Second Brain
Before starting, make sure you have:
An Obsidian vault: Any vault works - new or existing. Obsidian is free to download at obsidian.md.
Claude Code installed: You'll need a Claude Pro or Max subscription (around €20/month for Pro). Claude Code runs in your terminal.
Basic terminal comfort: You'll be typing commands in a terminal. Nothing complicated, but you should know how to navigate folders with cd.
Node.js installed: Claude Code requires Node.js to run. Download it from nodejs.org if you don't have it.
Step 1: Set Up Your Obsidian Vault Structure for AI
A second brain only works if it's organized in a way that both you and Claude can navigate. You don't need anything elaborate - a simple folder structure is better than a complex one.
Create the following folders in your vault:
daily/ - Daily notes organized by year (e.g., daily/2026/2026-04-13.md)
projects/ - One folder per active project
areas/ - Ongoing responsibilities (health, finances, career)
resources/ - Reference material, research, clippings
_inbox/ - Unprocessed captures and quick thoughts
templates/ - Note templates for consistency
This structure is inspired by Tiago Forte's PARA method (Projects, Areas, Resources, Archives) and works beautifully with AI because each folder has a clear purpose that Claude can understand.
Pro tip: Name your files like sentences, not codes. meeting-notes-product-redesign.md gives Claude useful context before it even opens the file. untitled-48.md gives it nothing.
Step 2: Create Your CLAUDE.md File — Claude's Persistent Memory
This is the most important file in your entire setup. The CLAUDE.md file sits at the root of your vault and acts as Claude's persistent memory - it's automatically loaded every time Claude Code starts a session in your vault.
Create a file called CLAUDE.md in your vault's root directory and add the following sections:
Who you are: Your name, what you do, what you're working on. Give Claude the context a trusted colleague would have after working with you for months.
Vault structure: Describe your folder layout and what goes where so Claude knows where to find and save things.
Writing rules: How you like your notes formatted. Do you use bullet points or prose? Tags or folders? YAML frontmatter or not?
Current priorities: Your active projects, deadlines, and focus areas. Update this regularly.
Conventions: Any naming conventions, tag systems, or linking patterns you follow.
Think of CLAUDE.md as writing an onboarding document for a very capable new assistant. The more specific you are, the better Claude performs. Don't write it for humans - write it for the model.
Step 3: Create a Memory File for Dynamic AI Context
While CLAUDE.md is your static context (things that rarely change), a memory.md file captures dynamic context - things Claude learns about you over time.
Create a memory.md file at your vault root. This file can include:
Preferences Claude has learned: "User prefers concise summaries over detailed breakdowns." "User always wants action items extracted from meeting notes."
Past decisions and reasoning: Why you chose one approach over another, so Claude doesn't suggest things you've already rejected.
Recurring patterns: Projects that come up often, people you work with frequently, tools you use.
The key insight here is that your context lives in your files, not in AI memory. It's readable, editable, and fully under your control. If Claude starts getting something wrong, you just edit the file.
Step 4: Connect Claude Code to Your Obsidian Vault
Now for the actual connection. This is surprisingly simple.
Option 1 - Direct launch: Open your terminal, navigate to your vault folder with cd, and type claude. That's it. Claude Code automatically reads CLAUDE.md and has access to every file in the vault.
Option 2 - Right-click launch (macOS): Find your vault folder in Finder, right-click, go to Services, and select New Terminal at Folder. Type claude and you're in.
Option 3 - From a separate project: If you're coding in a different folder but want Claude to have vault context, launch Claude Code with the --add-dir flag:
claude --add-dir ~/path-to-your-vault
From this moment, Claude can read your notes, create new ones, search across your entire vault, and make connections you never would have found manually.
Step 5: Automate Daily Notes with Claude Code Skills
Claude Code uses Skills - small instruction files that teach it specific workflows. Your first skill should automate your daily note.
Create this folder structure inside your vault:
.claude/
skills/
daily/
SKILL.md
Inside SKILL.md, write instructions for your daily note workflow. Tell Claude to:
Check if today's note exists at daily/2026/YYYY-MM-DD.md. If not, create it from your template.
Read yesterday's note and extract any incomplete tasks.
Check active project notes to surface current priorities.
Generate today's note with sections for Focus (top 3 priorities), Tasks (carried over + new), Notes (freeform), and End of Day (left empty for evening reflection).
Each morning, open your terminal in the vault and tell Claude to run the daily skill. Claude reads yesterday's note, pulls in unfinished tasks, checks your projects, and builds a ready-to-go daily note. Your morning starts informed instead of scrambling.
Step 6: Add an AI Research Skill to Your Obsidian Vault
The second most useful skill is automated research that feeds directly into your vault.
Create another skill at .claude/skills/research/SKILL.md with instructions like:
Search the vault first to check if notes on the topic already exist.
Search the web for current information.
Synthesize findings into a structured note with Overview, Key Concepts, Connections (links to existing vault notes), and Sources.
Save to resources/ with proper frontmatter and wikilinks to related notes.
Now you can say something like "Research the Zettelkasten method and save to my vault" and Claude does everything: finds what you already have, searches the web, writes a structured note, and links it to your existing knowledge. One prompt in, your vault is smarter.
Step 7: Set Up Automated Weekly Reviews with Claude Code
The real power of a second brain shows up when you review and connect what you've captured.
Create a weekly review skill at .claude/skills/weekly-review/SKILL.md that instructs Claude to:
Read all daily notes from the past week.
Summarize key themes, decisions, and accomplishments.
Identify patterns - what kept coming up? What got neglected?
List incomplete tasks that need to carry forward.
Generate a weekly summary note saved to a reviews folder.
Run this every Friday or Sunday. Over time, your weekly reviews become an incredible record of what you've done, how your thinking evolved, and what patterns repeat.
Advanced: AI Ingestion and Knowledge Synthesis in Obsidian
Once your basic setup is running, you can expand into more powerful workflows:
Voice memo ingestion: Record a voice memo after a meeting, and Claude can transcribe it, identify speakers, extract decisions and action items, and distribute them across the relevant notes in your vault.
Web content ingestion: Feed Claude a URL and it doesn't just summarize into one note - it can update existing pages, resolve contradictions with what you already know, and create new connection notes.
Whiteboard capture: Take a photo of a whiteboard, and Claude extracts text and structure, creates concept notes, and links them to related projects.
Decision challenging: Before a big decision, ask Claude to search your vault for past failures and reversed decisions on the same topic. It pushes back with your own words. Your vault holds you accountable.
Tips for Getting the Most Out of Your Obsidian AI Second Brain
Be specific with prompts: "Summarize my work" gives Claude nothing to work with. "Summarize all notes tagged #client-work from the last two weeks and list open deliverables" gives it everything.
Update CLAUDE.md regularly: As your vault evolves, your context file goes stale. Review it monthly.
Use consistent frontmatter: YAML frontmatter with tags, project names, dates, and status fields makes it dramatically easier for Claude to find and filter your notes.
Use wikilinks liberally: Obsidian's [[wikilinks]] create a graph that Claude can traverse to understand how your concepts relate. The more links, the smarter the system.
Start small: Don't try to automate everything on day one. Begin with daily notes, get that workflow reliable, then add research, then weekly reviews. Let the system grow organically.
Review Claude's output: Claude writes to your files. Always review what it creates before treating it as your own thinking. It's an assistant, not an oracle.
Limitations to Keep in Mind
Context window: Claude Code still has a context window. Very large vaults with hundreds of notes require targeted queries - Claude can't read 800 notes in one session. Structure your prompts to point it at specific folders or tags.
Terminal requirement: This setup requires comfort with the command line. If you've never used a terminal before, there will be some learning curve during initial setup.
Cost: Obsidian is free, but Claude Pro runs approximately €20/month. That said, this is significantly cheaper than most AI productivity tools that charge subscriptions plus per-token API costs.
Not real-time: Claude doesn't run in the background watching your vault. It acts when you prompt it. For scheduled automation, you'll need to set up cron jobs or similar triggers.
Summary
Building an AI second brain with Obsidian and Claude Code comes down to a few key steps:
Set up a clean vault structure that both you and Claude can navigate.
Create CLAUDE.md as Claude's persistent context file - this is the heart of the system.
Add a memory.md file for dynamic context that evolves over time.
Connect Claude Code by launching it inside your vault directory.
Build skills starting with daily notes, then research, then weekly reviews.
Expand gradually into ingestion, synthesis, and decision-support workflows.
The system compounds over time. The more notes you take, the more connections Claude can surface. The more you use it, the better it understands your thinking patterns. Your vault stops being a place where ideas go to die and becomes a place where ideas evolve.
Your notes are the moat. Claude is the engine. Together, they build something neither could alone.
Happy building!