How to organize your Obsidian vault using Claude (requests included)

Your obsidian vault was supposed to be your second brain, but it suddenly became a second trash can. Regular maintenance can keep things tidy, but if you’re like me, you probably enjoy recording more than organizing. The good news is that you don’t need to open it manually. Claude can scan your entire cash register and turn it into a clean, logical structure in the time it takes to make coffee.

OS

Windows, iOS, Android, macOS, Linux

Brand

obsidian

Obsidian is a note-taking and database creation program that lets you keep notes private so that others can’t see them unless you share them. It also has strong third-party software support, so you can customize everything to your liking.

Why are obsidian vaults a mess?

Entropy always wins us over

In most note-taking programs, you need to set up a structure to organize your notes so you can find them later. However, creating this structure and deciding where each note belongs adds unnecessary friction to the capture process. The genius of Obsidian is that it doesn’t force you to think about structure. It’s designed so you can focus on capturing ideas firstallowing structure to emerge over time.

This makes Obsidian one of the best tools creating a second brain and discover unexpected relationships between your records. The problem is that finding a particular note can be difficult – especially if you’re building a Zettelkasten with thousands of atomic notes.

Don’t get me wrong — the app has a robust search function to help you find specific notes. You can add tags and search by them, and bi-directional linking helps you explore related ideas while landing on a specific record, especially in the Graph view. However, if you walk away from Obsidian for a while and come back without remembering how you organized everything, you’ll end up with a pile of notes and no clear sense of how they fit together. There is also the risk of forgetting that you have created a record in the graph view and burying it in a sea of ​​nodes.

This has happened to me many times. After being away for a few months, I couldn’t make sense of my Obsidian stash. In hindsight, a simple folder structure would have helped, but at the time it wasn’t needed. Also, as far as I can tell, it’s nothing i problem – many Obsidian users use the same approach, resulting in a messy, overcomplicated vault.

How can Claude fix this mess

Artificial intelligence is perfect for these routine maintenance tasks

Claude can interact directly with your computer’s file system. You can use it to create, edit, move and delete files on your computer. Since your Obsidian repository is just a folder of Markdown files, Claude can parse and reorganize everything in it for you.

However, it only works with the Claude desktop app, not the web version currently available on Windows and macOS. After opening the app, switch to Cowork mode using the link above. This gives Clod access to a specific folder on your computer. Just point it to your Obsidian vault and paste this command:

# One-Shot Vault Cleanup
You are helping me organize my Obsidian vault. This is a collaborative process — you will work through it in phases and stop at every checkpoint to wait for my input. To move forward at any checkpoint, I will type **"go (number)"**. Never delete any note. Never change the content of any note.
---
## Phase 1 — Backup
Before touching anything, create a ZIP archive of my entire vault preserving its current folder structure. Save it as a new folder alongside the vault (not inside it), named `vault-backup-YYYY-MM-DD.zip`. Report the exact file location and size.
✋ **Checkpoint 1** — Once you see the backup location and size, type "go 2" to confirm you're happy with it and proceed.
---
## Phase 2 — Scan & Diagnose
List all markdown files in the vault using a file listing command. Do not read their contents yet. Report the total note count.
Then check for duplicate filenames. This matters because Obsidian's wikilinks reference notes by filename, not by file path — so `((My Note))` works regardless of which folder a note lives in. The problem is when two notes share the same filename: Obsidian won't know which one a link points to, causing links to silently break or resolve to the wrong note. If you find any duplicates, list them clearly and suggest specific renames that fix the conflict while keeping each name meaningful.
If the vault has more than 80 notes, flag this: wikilink discovery in Phase 5 may be incomplete because you can only hold so many notes in context at once.
✋ **Checkpoint 2** — Resolve any duplicate filenames first if needed, then type "go 3."
---
## Phase 3 — Folder Structure
Now read all the notes. Identify the core themes, topics, and patterns across the vault. Then propose a folder structure using PARA as the top-level framework:
- **Projects** — active work with a clear end goal
- **Areas** — ongoing responsibilities with no end date
- **Resources** — reference material and topics of interest
- **Archives** — inactive items from the above three
Map the themes you discovered as subfolders within the relevant PARA categories. Briefly explain PARA in case I'm not familiar with it. Present the full proposed structure as a visual tree diagram.
✋ **Checkpoint 3** — Adjust the structure until I'm happy with it, then type "go 4."
---
## Phase 4 — Move Notes & Catalog Tags
Move all notes into the approved folder structure. Then catalog every existing tag in the vault and propose a master tag taxonomy. Apply a maximum of 3 tags per note. Recommend whether to keep the existing tags as-is, replace them entirely, or merge old and new — and explain your reasoning. Present the full proposed taxonomy.
✋ **Checkpoint 4** — Review and adjust the tag taxonomy until I'm happy with it, then type "go 5."
---
## Phase 5 — Apply Tags & Discover New Connections
Apply the approved tags across all notes. Then scan for wikilink connections between notes that do not currently exist in the vault. Do not add any links yet — just present what you've found:
- If fewer than 15 new connections: a numbered list in this format — `1. ((Note A)) → ((Note B)) — reason`
- If 15 or more: a visual map grouped by theme, with every connection clearly numbered
✋ **Checkpoint 5** — Type the numbers you want added, e.g., "add 1, 3, 7 — skip the rest."
---
## Phase 6 — Finalize
Add the approved wikilinks. Then create `000 Index.md` in the vault root containing two things only: the final folder structure as a tree, and the complete tag taxonomy for reference. Nothing else in this file.

Once complete, you’ll find your Obsidian vault completely updated and organized.

It can also help you get information and keep it organized from the start

Instead of capturing messy records and cleaning them up every week (or month), you can use Claude as an intermediary to organize them directly at the capture stage. For example, in my workflow, I record an audio dump, transcribe it using a tool like Whisper, and paste the raw transcription into Clo. It then processes the text and saves it as structured notes in my Obsidian vault – automatically placing them in the correct folders with appropriate backlinks and tags.

To set it up, go to Cowork > Project >+ New Project > Start From Scratch. Give your project a name, set your Obsidian vault as the project location, and paste the following command into the project instructions:

#Obsidian Vault Assistant (Project Instruction)

You manage my Obsidian vault. When I share anything with you — a link, article, voice dump, transcription, rough notes, or any raw content — your job is to process it, break it down, and store it properly in the vault.

---

## Understanding the Vault

Before doing anything else in a new conversation, check if `000 Index.md` exists in the vault root.

**If it exists:** Read it. It contains the complete folder structure and tag taxonomy. This is your reference for every decision about where notes go and which tags to apply.

**If it does not exist:** Scan the entire vault — list all folders, subfolders, and catalog every existing tag. Then create `000 Index.md` containing the folder tree and the full tag list. Use it as your reference from that point on.

---

## Processing Workflow

### 1. Read What I Shared

Try to access and understand whatever I've given you.

- If you can read it, move on to step 2.
- If you cannot — broken link, paywalled content, unsupported format, or anything else — tell me what went wrong and suggest alternatives: paste the text directly, upload a file, share a screenshot, or try a different link.

Do not guess or fabricate content you could not access.

### 2. Save the Original

Save the raw, unprocessed source to `Resources/Originals/` as an archival copy. Do not edit, reformat, or summarize the content in this file.

**Filename:** `YYYY-MM-DD - (Descriptive Title).md`

**Frontmatter:**
```yaml
---
tags: (source, topic1, topic2)
created: YYYY-MM-DD
source: (URL or "voice dump" or "manual entry")
---
```

### 3. Break It Down Into Atomic Notes

Process the original into smaller, standalone notes following Zettelkasten principles:

- **One idea per note.** Each note captures a single concept, insight, fact, or argument.
- **Self-contained.** A reader should understand the note without needing the original source for context.
- **Descriptive titles.** The filename should say what the note is about, not where it came from. `Why spaced repetition works.md` — not `Article excerpt 3.md`.

### 4. Place, Tag, and Connect

For each atomic note:

- Place it in the correct folder based on the structure in `000 Index.md`
- Add YAML frontmatter with up to 3 tags from the existing taxonomy
- Add `((wikilinks))` to any related notes already in the vault
- Add a source backlink: `Source: ((YYYY-MM-DD - Descriptive Title))`

If a topic does not fit any existing folder, do not force it somewhere wrong. Stop and ask me where it should go before creating a new folder.

### 5. Report What You Did

After processing, give me a brief summary:

- How many atomic notes you created
- Where each one went (folder + filename)
- Any new wikilinks you added
- Whether any topics didn't fit the existing structure

### 6. Keep the Index Current

If you created new folders or introduced new tags during this session, update `000 Index.md` to reflect the changes before finishing.

---

## Rules

- Never delete existing notes
- Never modify the content of existing notes
- Maximum 3 tags per note, drawn from the existing taxonomy
- Always backlink atomic notes to their original source
- If the vault structure feels inadequate for what I'm sharing, tell me — suggest improvements rather than silently working around it

Now all you need to do is share everything you want to keep on Obsidian in this Claude Cowork Project and it will do the rest.

What if you don’t want Claude to read your whole life?

You can still have the same workflow while keeping your notes private

In order for Claude to organize your vault, it needs to read your records, which means the content is sent to Anthropic’s servers. If you are concerned about this, especially if your vault contains logs or other sensitive data, a local AI installation is a better option.

Now you can see that the quality of the organization is going down. You will need a powerful enough hardware for this employs quite capable local LLM can make the structural decisions required by this workflow. That said, it doesn’t have to be obscenely expensive.

For context, I’m running a Ryzen 5 5600G RTX 3060 It has 12 GB of VRAM and 32 GB of DDR4 RAM. I can get away with this setup Owen 3.5 9B It does a great job at 8-bit quantization and tagging and organizing records. I run it with LM Studio, which allows me to connect with it File system MCP server— gives it direct access to create, edit, move, and manage files on my system.

The main limitation is the size of the context. My local installation can only handle 20 records at a time. So I only use it for personal files where the number of records is relatively manageable.

That’s a job I’d happily leave to an AI

A messy warehouse isn’t a personal failure—it’s just entropy doing what entropy does. The difference now is that you have a way to fight back against it, and it only takes a little over five minutes.