01What a Claude Second Brain Obsidian Setup Is

A claude second brain obsidian setup is an Obsidian Vault full of notes about your business that Claude can read before it writes. It means Claude remembers your voice, your methods and your real customer stories, instead of giving you a generic answer.

Obsidian is a free app for your notes. Every note is just a plain text file in a folder on your computer, and that folder is called a vault. You link notes together with double square brackets.

Type two brackets, type a note's name, and the two notes are now connected.

Open the graph view and you see every note as a dot and every link as a line. Julian Goldie calls it his business as a map.

So why does Claude need this? Every Claude Code session begins with a fresh context window, which Anthropic's own docs confirm. Every chat starts from zero.

Claude can read plain text files, so when you point Claude at the vault folder, it can read everything your business knows.

In this episode, Julian shows three systems working together:

  • The Obsidian Vault, which holds what your business knows.
  • The guide writer skill, which tells Claude how to write a guide.
  • Claude memory, the rule book Claude follows every single session.

02What Goes Inside the Vault

Julian's vault is small and simple. It holds a handful of note types, and each one has a clear job. The key is that every note links to the others, so Claude can follow the trail from one idea to the next.

  • Brand voice: how the business sounds when it writes.
  • Words We Avoid: a note listing words like leverage and robust. Claude swaps them for plain words.
  • SOPs: how things get done. Examples include the guide writer SOP, onboarding a new member, the weekly SEO report and answering community questions. Each one is a short list of steps.
  • Stories: real member stories. Julian calls these the gold.

The links matter. For example, the SEO report note links to Kirk's story. When the links connect, the graph view lights up.

Julian's demo vault had 39 notes. That is all it takes to start, and you don't need a thousand. Because everything is just files, nothing is locked in.

You can open the same notes in any text editor, and you can back them up. Claude can also read them, search them and write new notes straight into the vault.

03How to Do Your Claude Second Brain Obsidian Setup

Julian walks through the setup from zero. The steps below are the ones he spells out in the episode. Before you start, have a vault folder ready with a few notes in it, such as your brand voice, one SOP and one real customer story.

  1. Download Obsidian. It is free.
  2. Open Claude Code in the same folder as your vault. In the desktop app, pick the vault folder as the project. In the terminal, go to the folder and type claude. Now Claude can read every note.
  3. Add your rule book. Make a file called Claude.md in the vault.
  4. Start it with four rules: read the relevant notes before you write, write in the brand voice, never invent a story or a number, and link every note you use.
  5. Add the guide writer skill. Make a folder called .claude, then a folder called skills inside it, then a folder called guide-writer, and put the skill.md file inside.
  6. Ask Claude for your first guide.

That is the whole setup. Once Claude is opened in the vault folder, it can search your notes, read your rules and use your skill without you pointing at each one by hand.

04How the Guide Writer Skill Works

A skill is a file where you write instructions, and Claude adds it to its toolkit. Claude uses a skill on its own when it is relevant, or you can call it by name with a slash.

Julian's guide writer skill sets a clear shape for every guide:

  • 5 to 7 steps.
  • Each step is something you click or type, with the exact prompt.
  • A real result section with the member's name and their own words.

Then he runs it. He types one line: write a guide for my community on setting up your first SEO agent. He doesn't even mention the skill.

  1. About 9 seconds in, Claude loads the guide writer skill by itself.
  2. It reads the brand voice note, the Words We Avoid note and the SOP.
  3. It searches the vault for SEO and finds the weekly SEO report SOP and Kirk's story.
  4. At 27 seconds, it writes the file.
  5. At 35 seconds, it is done.

This is the point of the skill. You give Claude the format once, and every guide comes out in the same useful shape without you repeating yourself.

05With the Vault vs Without It

Julian runs the same prompt twice to show the difference. Same Claude, same prompt. The only change is whether Claude has the second brain.

Without the vault, Claude writes a fine guide. It says an SEO agent is an AI that does your repeat SEO work. It is decent, but it could be anyone's guide.

There is no story, no proof and none of Julian's voice. It doesn't know his tools, his stories or how he talks.

With the vault, the guide is very different:

  • It opens with the result: you end up with an agent that checks your SEO every week.
  • It has 6 steps, and each one has the exact prompt to paste.
  • It points to Hermes agent and Gemma, the tools that are in Julian's vault.
  • It includes Kirk's real result. Kirk had been dabbling in basic AI use, and now he is building an agent OS, taken word for word from his review.
  • At the bottom, it links to every note it used.

The vault version sounds like Julian and backs it up with a real person. That is the whole difference a second brain makes.

06Claude Memory: Claude.md and Auto Memory

Here is the problem with any AI. You correct it, it gets it right, and then next week it makes the same mistake again. Claude Code has a fix for that: a file called Claude.md.

Claude reads it at the start of every session. Anthropic's docs say to add to it when Claude makes the same mistake a second time.

Julian's vault has one at the top. It is a rule book with rules like these:

  • Write in the brand voice.
  • Use real member stories.
  • Never invent a story, a name or a number.
  • Use UK spelling.
  • Link every note you use.

There is a second kind of memory too. Claude has auto memory. These are notes Claude writes itself based on your corrections and preferences.

Both load at the start of every session. Claude.md is the rules you write. Auto memory is what Claude learns on its own.

Together, they make the rule book that keeps Claude in line from one session to the next.

07Turning One Correction Into a Permanent Rule

Julian shows how the rule book grows. After reading the SEO guide, he wants one change. Every guide must end with one line that starts with do this today and gives one action.

He asks Claude to add that as a rule so it never forgets.

  1. Claude opens Claude.md.
  2. It adds the rule that every guide must end with a single final line starting with do this today.
  3. It fixes the SEO guide.
  4. It is done in 23 seconds.

Then comes the real test. Julian starts a brand new session, with nothing carried over from the chat. He asks for a different guide: how to answer member questions faster with Claude.

Claude loads the skill and reads the voice. It finds the SOP for answering community questions. This time it picks Thanh's story and Jeremy's story, because they are tagged support.

And the last line starts with do this today, telling the reader to take the last member question, run steps one to three and post the reply.

Julian said the rule once. That is the loop: every time you correct Claude, the fix becomes a rule, and Claude gets better every week.

08Tips and Mistakes to Avoid

Julian shares three tips that make a Claude second brain work. They are simple, but skipping them is where most setups go wrong.

  1. Keep stories real. Copy the customer's exact words and note where they came from. Claude will quote them, so they need to be true.
  2. Tag everything. Tags are how Claude picks the right story for the right topic. That is why Thanh's and Jeremy's support stories showed up in the support guide.
  3. Add the rule, not just the fix. Every time Claude gets something wrong, don't only fix the output. Add a rule to Claude.md so it doesn't happen again.

A few more things to keep in mind:

  • Start small. 39 notes was enough for Julian's demo.
  • Keep a Words We Avoid note so Claude drops filler words.
  • Link notes together so Claude can follow the trail from an SOP to the right story.

The biggest mistake is treating each chat as a one-off. Without a vault and a rule book, Claude starts from zero every time, and you end up correcting the same things again and again.

09Who This Setup Is For and Why It Matters

This setup is for any business owner who uses Claude to write and is tired of generic answers. If you run a community, an agency or an online business, your real value is in your voice, your methods and your customer stories. A vault puts all of that where Claude can find it.

It is not just for guides. The same vault works for:

  • Emails.
  • Onboarding messages.
  • Scripts.
  • SOPs.

Any time Claude writes for your business, it writes from your brain. And because the notes are plain files, you own them, you can back them up and you can open them anywhere.

Julian notes that his exact vault setup, with the folders, note templates, guide writer skill and Claude memory rules, is inside the AI Profit Boardroom, along with the agent OS. The main idea works for anyone, though. Give Claude a second brain and it stops forgetting your business.