---
title: "Obsidian + Claude Code: Your Second Brain, Amplified"
author: "@heyamit_ (with Vin)"
description: "This is Obsidian, a little tool that people are using as their second brain. What is really cool about it is that they are pairing it with Claude Code and getting crazy results. It is literally a..."
source: "https://adhenawer.net/posts/original/obsidian-claude-code-second-brain-amplified.html"
lang: en
---

TL;DR · key points

1. This is Obsidian, a little tool that people are using as their second brain.
 2. I have been slow to adopt Obsidian because it has been a little daunting to look at.
 3. People who understand how to use Obsidian and Claude Code together are going to be able to live happier, healthier, and wealthier lives.

## INTRODUCTION: A GAME-CHANGING PAIRING 0:00

This is Obsidian, a little tool that people are using as their second brain. What is really cool about it is that they are pairing it with Claude Code and getting crazy results. It is literally a game changer.

I have been slow to adopt Obsidian because it has been a little daunting to look at. So I brought my friend Vin, also known as Internet Vin, to explain what Obsidian is, how to use it with Claude Code, and how to set up commands that really drive the most out of Claude and all the LLMs.

People who understand how to use Obsidian and Claude Code together are going to be able to live happier, healthier, and wealthier lives. Why? Because it gives you incredible ideas on tap, at the right time, at the right moment.

By the end of this conversation, you will understand how to use Claude Code and Obsidian as a thinking partner, how to stop having to explain things to agents over and over again by simply passing specific files in, and how to use these tools to notice things about the way you think that you would not have noticed on your own.

## WHAT CLAUDE CODE IS, AND WHY CONTEXT IS THE WHOLE GAME 9:58

Claude Code is an agent you use in a command line interface. It is a tool that can control your computer through natural language. You can say “make a file on my desktop that says hello Greg in plain text,” and it goes and does it. Before this, you had to go to the desktop, open a text editor, and create the file manually.

What is interesting is that if you have an agent that can control your computer, then whatever you can describe to it, it can start to do. The more information it has, the more complex things it can do.

But there is a problem. If you write a long project description, or have an hour-long conversation with the agent, you do not want to explain all of that again in a new session. A lot of people use Claude or ChatGPT on the web with memory, but you cannot control what is in that memory. You do not know what it knows and what it does not.

So there needs to be a way of passing information into these agents that is easier and faster. The whole game is feeding the beast good context. The better and faster the information you can give it, the more it can do for you.

The solution is a file. If I write a project description, save it on my desktop, Claude Code can reference that file and pull it in whenever I want. In a new session, instead of explaining the project again, I just point to the file. That saves a lot of time.

This is the problem a lot of people face with Claude Code. They say it is “okay, not game changing.” The issue is they are not feeding the right context at the right time.

## WHAT OBSIDIAN IS: A GRAPH ON TOP OF MARKDOWN 10:09

Obsidian is an interface that sits on top of a collection of Markdown files. A daily note is just a Markdown file. A note on a concept is just a Markdown file. The whole collection is called a vault.

What makes a vault different from a folder is that Obsidian lets you make interrelationships between files. If I write “today I am on a podcast with Greg Eisenberg,” that file is now linked to the Greg Eisenberg file. People love Obsidian because of this ability to form interrelationships. A regular folder on your computer cannot show those connections.

It gets really interesting when you keep making connections over time. In the graph view, each circle is a file, and you can see how everything is connected. A note on “personal agent infrastructure” links to agentic AI, to Telegram, to Toby from Shopify, to a podcast. Each file also has a local graph showing every time you wrote about that subject.

The reason people love Obsidian is because it works more the way your brain works. Your brain connects patterns all the time.

## THE OBSIDIAN CLI: GIVING CLAUDE CODE YOUR SECOND BRAIN 10:14

Obsidian released a new tool called Obsidian CLI. It lets Claude Code read all the files in your vault, and crucially, it also gives Claude Code information about the interrelationships between those files.

Claude Code can see that this file is connected to that file, and that file, and that file. That becomes very interesting in terms of what Claude Code can understand about you and about the relationships between the things you are working on.

It can start to surface patterns about what you are thinking about that you are not seeing for yourself. Some idea you might have been writing about for a year could be a latent idea, and it can immediately say, “Hey, did you know you have been writing about this same pattern in startups, or in this project, across every single note you make in different domains?”

Seeing that for the first time can be like a huge light bulb. It can cause huge progressions in your learning, your understanding, your point of view on the world, and in what you are working on.

## CUSTOM COMMANDS: LOADING CONTEXT IN ONE KEYSTROKE 12:25

I have a set of slash commands I created. You do not need to be afraid of them. I run them from a terminal tab inside Obsidian, but you can run them from any terminal. You can also ask Claude Code to create these commands for you, just by describing what you want.

/context — Loads full context about my life, work, and current state. It reads context files, daily notes, and follows backlinks to build a complete picture. Before I start working on something, I type “/context demo” and it preloads everything it needs to know.

/today — Morning review. Pulls calendar, tasks, iMessages, and the past week of daily notes into a prioritized plan for the day. An agent with only calendar and iMessage access does not know what you are thinking about or why. If I have been writing daily notes about a particular project, do my calendar and messages reflect that? With this context, it can more effectively decide what should or should not be on my calendar.

/close-day — End-of-day processing. Extracts action items, surfaces vault connections, checks confidence markers. I keep hypotheses with confidence ratings: an idea I feel very solid about, another I am still unsure of.

These are daily operational commands. What I use Obsidian for the most is thinking tools.

## THINKING TOOLS: USING THE LLM AS A THINKING PARTNER 14:51

I really like using LLMs as a thinking partner. That is my favorite way to use them. Some people use agents to build things; I like to think alongside them.

/ghost — Answers a question the way I would. It builds a voice profile from the vault, writes in that voice, then evaluates the fidelity.

/challenge — Pressure tests current beliefs using the vault's own history. Finds contradictions, counter-evidence, and shifts in thinking. If I want to keep developing, I need to make sure my point of view is not overly biased or limited.

/emerge — Surfaces ideas the vault implies but never states. Conclusions from scattered premises, unnamed patterns, unarticulated directions. A lot of times I have been circling an idea for years, and someone simply naming it for me is a huge breakthrough.

/drift — Compares my stated intentions against actual behavior over 30 to 60 days. Surfaces what I am avoiding.

/ideas — Deep 30-day vault scan with cross-domain pattern detection and graph analysis to generate ideas across all domains. It does not only give me ideas for things to work on; it also gives me ideas for films I should watch, products I should buy, all influenced by what I am writing about.

/trace — Tracks how an idea has evolved over time across the vault.

## AN EXAMPLE OF TRACE: WATCHING AN IDEA EVOLVE 18:14

I ran “/trace” on my own relationship with Obsidian. The agent said: “I will trace how your relationship with Obsidian has evolved. Let me start by building a vocabulary map and searching across the vault.” It read files, followed links, and produced a full evolution trace.

Pre-vault, December 2024: an essay titled “How I take notes in the West End of Toronto” describes a complete system where Obsidian plays no role. The pipeline is capture, process, crystallization, using MacWhisper, LLM dialogue loops, Canopio for spatial mapping, physical notebooks for cognitive friction, Futureland for tracking, Are.na for fragments.

Discovery and skepticism, January to May 2025. The first daily note in the vault: raw excitement mixed with uncertainty. “I could probably drop transcriptions in here too. The current theory is that bidirectional linking is not that useful, but I do not know.”

A pivotal realization on backlinking: “Originally I was backlinking to general terms, podcasts or physical fitness or filmmaking. I am realizing it is not the most useful way to use Obsidian. The most important thing is to create notes for each of my patterns, theories, projects, or perspectives, and link to those.”

Phase four, January 2026, a month of explosive building. “Everything still requires me to actively prompt and manage each session. The next unlock is figuring out how to get agent tasks to run automatically. The friction is no longer with Obsidian itself, but with the boundary between the vault and agent execution.”

This is a pattern I would not be able to see on my own, or this fast. A computer can have this much information about me and surface these patterns. I could do this with startups, a project, foods, relationships, a hobby, parenting, anything.

## MAKING REFLECTION A HABIT 22:52

Part of getting good at Obsidian is inserting reflection into your everyday life. A lot of people move from meeting to meeting, busy, parents, and as we get older we actually write and reflect less.

There are two reasons reflection matters to me. First, it is amazing to look back. Seeing that I am a person continually changing, my skill continually changing, projects continually evolving. Being able to reflect on how things change over time is an amazing part of life.

Second, there is a functional reason. Writing is how I generate ideas. When I sit down at the computer and write things down, ideas come. By writing something out, I internalize it more. If you want to cultivate writing as a habit, connect it to the idea that this is how you progress, this is how you generate original ways of thinking.

Writing is also how you delegate to agents. That is a whole new aspect. If you develop a writing habit, you have much more context to pass to an agent, which dramatically increases what you can delegate and what you can build.

## FROM MANUAL PROMPTING TO AUTONOMOUS AGENTS 25:50

OpenCloud is an autonomous agent that can go and do things without you prompting every time. It makes decisions and builds things on its own.

Here is how that connects. I have a /schedule command. I asked it, “Can I take a meeting with Greg Eisenberg on Feb 20 at 2pm?” It looked at my calendar, my daily notes, what I care about, and responded: “Your day is stacked. You are already recording Greg's podcast this morning, followed by a team lunch and a meeting with Peter and Vince. Your Feb 17 notes show the Greg episode has been top of mind. The vault has a dedicated Greg Eisenberg note. Recommendation: not 2pm, and you might not need a separate meeting at all.” That is the correct answer.

An autonomous agent can do the same thing on its own. It can read my vault, find connections, and make decisions on my behalf with a deeper understanding of me. Instead of managing an agent or talking to another human, I just focus on managing this vault. The vault is the new source. I continually try to make sure it has all the information needed so I can delegate. If the agent is not making the right decisions, I change something in the vault, not in my prompts.

The scary part: if Obsidian is really your second brain, giving an autonomous agent access to it is scary. I have purposely given Claude Code access to a lot of personal information, because I want to understand what these things are and how our relationship with computers is changing. But you have to think carefully about how much you are sharing. Even for this demo, I had to create demo versions of every command so they would not reveal too much personal information on screen. Privacy as a concept is going to evolve in interesting ways.

## CONNECT: BRIDGING DISTANT DOMAINS 29:12

/connect — Takes two domains and connects them using the vault's link graph. I tested it with “filmmaking” and “world-building.”

It read notes in the filmmaking neighborhood (a 35-film watchlist, a first meeting with Toby), and in the world-building neighborhood (an essay on world-building, notes on my media company). Then it produced bridges.

Bridge one: the interview portal and the constructed world in filmmaking. “If I noticed something specific and asked a question about it, it would open a portal into a person's internal world, which is often a vast universe of concepts and beliefs and visions.” And from my world-building essay: “I want my blog to show you what I value, what I believe, what I worry about, like a tomb from ancient Egypt. I want my blog to be a place that you dig up and examine long after I am gone.”

Bridge two: always-on documentary equals continuous world-building. A creative strategy where companies continuously narrativize their characters, pursuits, conflicts, and visions.

I could connect shawarma and startups if I wanted to. The depth depends on what you are writing about. And none of it requires explaining anything to the LLM. I just type “/connect filmmaking world-building.”

## USING THE VAULT FOR MEETINGS AND EXTERNAL INPUT 32:00

You can use vaults however you want. If you want to drop Granola meeting transcripts in, create a “Meetings” folder, then a file for “Meeting Greg Eisenberg plus Vin,” paste the transcript. Now it is in the vault. You can pass it to the agent, or the agent will discover it, especially if you tag it back to a podcast or project. The agent now knows the transcript is related to that other file.

How much you put in is up to you. You might even want separate vaults for separate purposes.

For me, I use LLMs to increase my own level of understanding, so I use it for a lot of reflection. I do not want the agent writing into my files. I have a strict separation: I write my side, the agent writes its own side, and I take from it what I think should be included. Otherwise, when it finds patterns, is it finding patterns about what it has written, or patterns about what I have written? I want to keep that clear.

That said, AI is very good at going to the internet, finding information based on trends, distilling it the way you want, and putting it in your world. That is also interesting. If you ask Claude Code to go through your vault and generate ideas for tools you should build, you can say, “Great, now build the tool.”

## FROM REFLECTION TO BUILDING: THE IDEAS COMMAND 35:31

To move out of the realm of reflection and into building, I ran /ideas. These commands take a while because they read so many files. A big difference when using Claude Code with Obsidian access is that requests take longer, because it is pulling from way more material.

It ran through “obsidian-orphans,” “obsidian-dead-ends,” “obsidian-resolved,” tag counts, daily notes, and full context files for my projects. I maintain extensive context files. One of them documents what is shifting recently in my show, core beliefs of the show, research as the foundation, the team working on it. Another is a personal workflow context with my daily schedule and how I like to work. Another is a personal agent infrastructure file about how I am thinking about delegation.

All of that gets factored in to the task of generating ideas. It takes five minutes, but that is exactly what I want. I want responses from LLMs that are deeply contextual to what I am writing about. That is how an agent and I work best together: I focus on continually noting where I am with projects, what my current understanding is, what I find interesting. The agent always has the best representation of who I am in the moment.

The quality of information the agent has entirely determines what it can do for you. If it does not know much about you, it cannot do much for you. If it knows a lot, it can do things that, in a strange way, you do not even know about yourself.

What this is doing, at its core, is connecting the dots. It is quite difficult, as a business owner or in personal life, to connect your own dots. That is why people go to coaches and therapists. In therapy you do most of the talking, and the therapist guides you toward connections. Obsidian plus Claude Code does something similar: it helps you uncover what your dots are and how to connect them.

## THE IDEA REPORT: A CONCRETE OUTPUT 41:47

The /ideas report came back with structural highlights, working patterns, tools to build, tools to use, systems to implement, subjects to investigate, things to write, and conversations to have.

Structural highlights flagged orphans worth noting (defense technology in Canada sitting in isolation, agentic software), unresolved links that reveal latent interests, hidden relationships.

What is working: “Obsidian plus Claude Code as a combined system is producing genuine breakthroughs in thinking and output. Day-per-domain structure, when enforced, works. The Greg Eisenberg episode as a forcing function compressed months of thinking about Obsidian and agents into a clear thesis with demos.”

Tools to build: a /graduate command that scans recent daily notes, identifies tagged or untagged ideas, and prompts me to promote them into standalone notes, turning a daily note stream into a structured idea pipeline. An Obsidian vault for my team, so we can all contribute and ask it questions. A time-blocking app that enforces day-per-domain.

Systems to implement: “One sentence in Obsidian, agent handles the rest.” Write “schedule a call with person about topic this week” in a daily note, and have an agent pick it up and handle it. Inline delegation, potentially a new UX pattern.

Subjects to investigate: Christopher Alexander's pattern language applied to digital spaces. Black Mountain College as a model for physical programming. Authorless media as a concept. How Shenzhen's hardware ecosystem actually works.

Things to write: context architecture essay. The computer as a place. Software becoming fashion.

Conversations to have: real people it surfaced from my notes. The workshop host about technical programming. Steph Ango, Obsidian's CEO, about the vault as a place.

The fact that this is plain text, with no images, is kind of the point. I like the stripped-down aesthetic. If you prefer, you can say “turn this into a beautiful HTML file on my desktop,” and it will do it.

## AGENT-SUGGESTED COMMANDS: MOVING UP IN ABSTRACTION 47:16

I started by building commands myself. Then I asked the agent, “What commands do you think would be interesting?” Another pattern I like is moving up in abstraction: instead of saying “make a command that tells me what to focus on each day,” I say, “Based on my Obsidian vault and what you know about me, form an understanding of my skill level and the projects I am working on, and suggest commands that would take me from where I am to a higher level.”

The agent suggested /graduate. “Ideas, insights, and original thinking accumulate in detail in daily notes but rarely graduate into standalone notes where they can compound through backlinks. This command scans recent daily notes, surfaces the best candidates, and helps decide what to promote.” It described how it would scan, cross-reference the vault, present candidates, and graduate selected ideas into mini-essays with backlinks.

All of that text hits differently for me, because I know I am writing a lot about these things. Words like “mini essay” mean specific things to me. It is contextual. I told it to build the command, and now I just type /graduate.

I am going about making notes, and in parallel an agent is looking at my notes and giving me ideas on how to improve my workflow and my life. It can not only suggest, it can build.

## WHY MARKDOWN IS THE OXYGEN 50:29

If I were OpenAI or Anthropic, I would buy Obsidian. It is the missing link.

If you are serious about using LLMs to take your ideas and get the most out of them, if you are serious about building what people are calling a “personal OS,” and you are not using a centralized, Markdown-based note-taking tool, then you are not using LLMs properly. At the very least, you are not using them at their limit.

The difficulty is that it requires time to set up properly, and the UI is a daunting blank canvas. It does not say “write your preferences here.” You have to come up with these ideas yourself. But even when working with other humans, you have to find a way to explain things to them. The difference with agents is that you only need to explain once. Once it is in a file, you can always reference it.

A file is essentially a perfect memory. Human memory is unreliable. You can remember a haircut as the best of your life when it might have been the worst. A Markdown file is a fixed data point. It is only as biased as you were when you wrote it, at that moment in time.

People think tokens are the oxygen. They are not. Markdown files are the oxygen. They are the memories. What is a human being, the energy of it, or the memories? A bit of both, but in today's computer experience, MD files are severely underrated.

## A FUNDAMENTAL SHIFT, AND THE ALPHA 53:43

There is something going on here, some fundamental shift. None of us have the right vocabulary yet to explain it. We are figuring it out in real time.

Ninety-nine point nine-nine percent of people are not going to spend the time to set up something like this and make it part of their daily lives. The alpha — leading a more productive, happier, healthier, better, more money-making life — is in using something like this with an LLM.

This is not an endorsement of Obsidian specifically. Pick a tool. There is no excuse anymore for not writing down and reflecting into Markdown files, in a world where LLMs treat Markdown as oxygen.

I can just be making notes on my computer, like I have been since I was a kid, and suddenly an agent can scan through them, build things, and connect patterns I could never see. At the root of it, it is just a collection of interrelated Markdown files. And that might be all you need.