Claude Code can read files. Obsidian stores files. Put them together and you get an AI research assistant that knows everything in your vault --- every note, every web clip, every saved article.

This guide shows you how to connect Claude Code to your Obsidian vault and start asking questions across your entire knowledge base.

Why This Combination Works

Obsidian stores everything as Markdown files in a folder on your computer. Claude Code can read any folder you give it access to. When you point Claude Code at your Obsidian vault:

Claude can search across thousands of notes
It understands Markdown natively (headings, links, tags, frontmatter)
It can cross-reference sources and find connections you’d miss
It can write new notes and update existing ones
It has the full context of your knowledge, not just what fits in a chat window

This is the setup Andrej Karpathy described for AI research --- a personal wiki that an LLM can research across. Obsidian + Claude Code is the most practical implementation.

Step 1: Set Up Your Vault for AI

Your existing Obsidian vault probably works as-is. But a few organization tips help Claude work better:

Use descriptive filenames. meeting-notes-2026-04-12.md is better than untitled-3.md. Claude scans filenames when searching.

Use frontmatter. Tags, dates, and descriptions in YAML frontmatter help Claude understand what each note is about:

---
title: React Server Components Deep Dive
tags: [react, architecture, frontend]
date: 2026-04-10
source: https://example.com/article
---

Organize by topic, not date. Folders like research/, projects/, references/ are more useful to Claude than chronological sorting.

Step 2: Feed Your Vault Web Content With Save

The biggest limitation of most Obsidian vaults is that they only contain what you manually typed. Your best knowledge is scattered across the web --- articles, documentation, YouTube videos, tweets, Reddit threads.

Save is a Chrome extension that converts any webpage to clean Markdown. Save a page, and the Markdown file goes to your Save Vault folder. Symlink that folder into your Obsidian vault, and every web page you save becomes a searchable Obsidian note.

Save captures:

Articles and blog posts
YouTube videos (full transcripts)
Twitter/X threads
Reddit posts and discussions
Documentation and API references
Academic papers
Confluence and wiki pages

Each save is a clean .md file with proper frontmatter --- ready for Obsidian and Claude.

Step 3: Connect Claude Code

Open your terminal and start Claude Code in your vault directory:

cd /path/to/your/obsidian-vault
claude

That’s it. Claude Code now has access to every file in your vault. Ask it anything:

“What are the key takeaways from the articles I saved about distribution strategy?”
“Find all notes related to React performance and summarize the main patterns”
“What connections exist between my meeting notes from last week and the product roadmap?”
“Create a summary document of everything I’ve saved about pricing frameworks”

Step 4: Use MCP for Always-On Access

If you use Claude Desktop instead of Claude Code, set up the filesystem MCP server:

{
  "mcpServers": {
    "obsidian-vault": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/path/to/your/obsidian-vault"
      ]
    }
  }
}

Now Claude Desktop can read your vault on demand during any conversation. Save also has its own MCP server that gives Claude direct access to your saved web pages.

What This Looks Like in Practice

Monday: You save 3 articles about competitor pricing strategies using Save. They land in your Obsidian vault automatically.

Wednesday: You save a YouTube video transcript from a conference talk about go-to-market. It joins the collection.

Friday: You open Claude Code in your vault and ask: “Based on everything I’ve saved this week, what pricing strategy should we consider for our new product?”

Claude reads all 4 documents, cross-references them, and gives you an answer grounded in your curated research --- not generic training data.

Over time: Your vault grows to hundreds of notes and saved pages. Claude becomes a research assistant that knows your entire domain. The more you save, the smarter it gets.

The Compound Effect

This setup compounds. Each new note and each saved webpage makes Claude more useful. After a few months:

You have a searchable archive of everything you’ve read
Claude can answer questions by synthesizing across your entire collection
Your Obsidian vault becomes a genuine second brain --- one that an AI can actually read
You stop losing knowledge to forgotten bookmarks and tab hoarding

Get Started

Install Save to start clipping web pages as Markdown
Set up the Save Vault folder inside or alongside your Obsidian vault
Open Claude Code in your vault: cd vault && claude
Start asking questions across your knowledge base

The best AI knowledge base is the one you actually build. Save + Obsidian + Claude Code makes it effortless.