---
title: 'NEW Claude Obsidian 2.0 Changes Everything'
source: 'https://youtube.com/watch?v=rUhahX27j6w'
video_id: 'rUhahX27j6w'
date: 2026-11-20
duration_sec: 531
channel: 'Julian Goldie SEO'
---

# NEW Claude Obsidian 2.0 Changes Everything

> Source: [NEW Claude Obsidian 2.0 Changes Everything](https://youtube.com/watch?v=rUhahX27j6w)

## Summary

This video introduces Claude Obsidian 2.0, a free, open-source plugin for Claude Code that transforms AI chat outputs into a persistent, linked knowledge base stored as markdown files in Obsidian. The presenter explains how the tool solves the problem of scattered AI conversations by creating a searchable, source-verifiable vault, and demonstrates its integration into their 'Agent OS' system for shared memory across multiple AI agents.

### Key Points

- **The Problem: AI Memory is a Filing Problem** [00:00] — AI answers get buried in closed chats, forcing users to re-explain context repeatedly. The solution is a tool that automatically files and organizes AI outputs.
- **What is Claude Obsidian?** [00:59] — A free, open-source project (MIT license) by developer Daniel Agrécy. It's a plugin for Claude Code that creates a knowledge base from plain-text markdown files, viewable as a linked map in Obsidian. Inspired by Andrej Karpathy's LLM wiki pattern.
- **How It Works Day-to-Day** [01:41] — Drop files into an 'inbox' folder, run the 'ingest' command, and Claude converts them into linked pages. Users can then ask questions, and answers come from their own notes with pointers to sources.
- **Key Changes in Version 2.0** [02:08] — Added source and claim ledgers for verifiable answers, made asking read-only to prevent accidental edits, and improved safety with careful write plans and crash recovery. No more automatic git commits.
- **The 15 Skills** [04:01] — Five core skills: vault setup, source ingestion, Q&A, saving answers, and health checks. Additional skills include canvas board creation, log rollups, and reference skills for formatting and 'think'. Skills are also available for Codex, Open Code, Gemini, Cursor, and Windsurf.
- **Integration with Agent OS** [05:03] — Claude Obsidian serves as the memory side of the presenter's Agent OS, allowing multiple agents (like Claude and Hermes) to share a single, linked memory vault, eliminating the need to re-explain context.
- **What to Put in Your Vault** [06:10] — Anything you'd otherwise re-explain: services, SOPs, client notes, research, articles, course notes. Can be shared with a team for a 'business brain'.
- **How to Start Today** [06:36] — Five steps: 1) Install Claude Code, Python 3.11+, and Obsidian. 2) Install the plugin (commands on GitHub). 3) Open an empty folder as vault and run 'wiki' command. 4) Add 3-5 documents to inbox and run 'ingest'. 5) Ask a question and check the source.
- **Honest Note and Big Idea** [07:19] — The vault is only as good as what you feed it; always verify sources. The core idea is that AI shouldn't start with a blank slate daily—Claude Obsidian provides a persistent, source-linked knowledge base.

### Conclusion

Claude Obsidian 2.0 offers a practical solution to AI's memory problem by creating a persistent, verifiable knowledge base. By integrating it into an agent OS, users can achieve shared memory across multiple AI tools, saving time and improving consistency.

## Transcript

Claude Obsidian 2.0 changes everything. Your AI doesn't have a memory problem. You have a filing problem. Every great answer ends up buried in a chat you'll never open again. Now a free tool files all of it for you and I run it inside my
agent OS. Today I'm showing you Claude Obsidian 2.0. You'll see what it is, what the big rebuild changed, and how it plugs into my agent OS. And at the end I'll give you a way to start today with just one folder. But first let's talk
about the real problem. You use AI every day. You get plans, answers, and research. Then the chat closes. Next week you open a new one and you explain everything again, who you are, what you're working on, how you like things
done. That drains your time and your energy. The knowledge is there. It's just scattered across chats, notes, and apps, and none of it connects. So here's the fix. Claude Obsidian is a free open source project from a
developer named Daniel Agrécy. It's under the MIT license, which means anyone can use it. It's a plugin for Claude code and it gives Claude one new job, keeping a real knowledge base for you. What does that mean? It's just a
folder of plain text files on your computer. They're called markdown files. You open that folder in Obsidian, which is a free notes app. Obsidian shows your notes as a map with every note linked to others. The idea comes from Andrej
Karpathy's LLM wiki pattern, a wiki that keeps growing every time you use it. Here's how it works day to day. You drop a file into a folder called inbox. It could be an article, a web page, or a document. Then you run the ingest
command. Claude reads the file and turns it into linked pages in your vault. Your original file stays untouched. Later you ask a question about your work and the answer comes from your own notes with pointers back to where it came from. So
it's not chat history. It 's not a normal notes app. It's a that Claude builds and keeps tidy for you. That's how you stop explaining yourself every morning. Okay, now the rebuild. Claude Obsidian
2.0 came out on July 30th. The project has kept moving since then. It's on version 3.0.1 now from late October, but 2.0 is where the big rebuild happened. So here's what changed. First, it keeps track of where
things come from. Version two added source and claim ledgers. In plain words, every claim in your vault keeps a pointer back to its source. So when you ask a question, you can check the answer against the original. That's the point,
answers you can check. Second, asking is now read only. When you ask the vault a question, Claude only reads. It doesn't change your notes by accident. The health check is read only too. It looks for dead links, notes with no links at
all, missing details, and empty sections, and it gives the same result part. We've got a complete 30-day Agent OS roadmap, and the full Agent OS zip
file is inside the AI Profit boardroom ready to install. The roadmap takes you through Agent OS one step at a time so you always know what to learn next. There are also four coaching calls every week. So if you get stuck setting up a
vault or connecting it to your agents, you can ask live. Everything you're seeing today fits right inside that setup. Link in the description or go to ai profit boardroom.com. Third, it's built to be safe. Rights follow one
careful plan, and it's built to recover if something crashes halfway through. Helpers can read sources and draft pages at the same time, but they only draft. One part checks the full plan and applies it, and it no longer makes
automatic git commits behind your back. So 2.0 isn't flashy. It's about trust. You can let an AI file your notes because you can see where everything came from. Inside Agent OS, that matters even more because many agents lean on
the same notes. Now the skills. There are 15. Five are the core. One sets up your vault or adopts a folder you already have. That part matters because it organizes your notes instead of replacing them. One
ingests your sources into linked pages. One answers questions from the vault. One saves an answer as a note, but only when you ask. And one is the health example, one builds obsidian canvas boards from your wiki so you get a
visual map of a topic or a project. Another rolls up the log of what changed so your vault keeps its own history. The last three are reference skills. Two cover obsidian formatting and obsidian bases and one is called think. And it
isn't only for Claude. The project also links these skills to Codex, Open Code, Gemini, Cursor, and Windsurf so you can teach another agent the same skills. Now let me show you how it fits in my agent OS. I use tools inside my agent OS
every day and Claude Obsidian is the memory side of it. My agent OS has a memory section. Every memory links to the others. When my agents learn something, they update it. I can see notes my agents updated just minutes
galaxy where every star connects to other stars and I sort my vault with Para, which is a simple way to organize notes into projects, areas, and more.
Here's one real example. I was in the Hermes section of my agent OS. I asked, knowing what you know about me, what are some personalized SEO keywords I could create content around? It looked through my obsidian vault and gave me ideas that
fit me. I didn't explain my business first. I didn't repeat myself. It already had the context. That's what agent OS gives you. Different agents, one shared memory. Claude, Hermes, whatever you plug in, all reading from
the same vault so nobody starts from zero every morning. So what could you put in a vault like this? In general, anything you'd otherwise again and again. Your services, your SOPs, which are your step-by-step
processes, client notes, research, articles you've read, notes from a course. Once they're in the inbox and ingested, they become linked pages you can ask about. And you can share the same vault with your team so their
agents work from the same notes. That's the business brain idea. One place that knows how you work. Now, how do you start today with one folder? Here's the simple version. One, make sure you have Claude code, Python
3.11 or newer, and Obsidian. Obsidian is optional, but it's how you see the map. Two, install the plugin. It's two commands and they're on the project's GitHub page. Three, open an empty folder as your vault and run the wiki command.
It shows you what it will create before it writes anything. Four, put three to five documents into the inbox. Pick things you already explain a lot. Then run ingest. Five, ask a question and check the source.
That's the habit I promised. Always look at where the answer came from. Then run One honest note, this isn't magic. A vault is only as good as what you feed
it. And an answer is only as good as the source behind it. That's why the pointers matter. Use them. So here's the big idea. Your AI shouldn't wake up with a blank slate every day. Claude Obsidian 2.0 gives you a vault that keeps its
sources, links your pages, and answers from your own notes. And inside Agent OS, every agent can use it. Right now, there's a window. Most business owners haven't figured this out yet. And that window won't stay open forever.
be on learning this tool. Inside the AI Profit Boardroom, you get daily tutorials, so there's always something new to learn. Four coaching calls every
week where you can ask live. A prompt library, so you're never starting from a blank page. And the member map to find other building with AI agents, plus the full agent OS zip file ready to install with a complete 30-day roadmap. Link in
the description or go to aiprofitboardroom.com. If you want the full process, the SOPs, and over 100 AI use cases like this one, join the AI Success Lab. It's our free community. Links in the comments and
description. You'll get all the video notes from this episode plus access to 95,300 members. The AI Success Lab is free to join. The video notes from this episode help you follow along with everything we
covered step by step, and you'll be learning alongside other people who are using AI every day. So if you want to keep learning, hit the link and come join us.
