The idea of a "second brain" has been around for years in the personal knowledge management community. The premise is straightforward: offload information from your head into a trusted external system so you can think more clearly and retrieve what you need when you need it. Obsidian became the tool of choice for a lot of people doing this seriously because it's local-first, markdown-based, and doesn't lock you into any vendor's ecosystem.

What changed recently is that Claude has become a genuinely useful intelligence layer on top of that system. Not just for answering questions, but for helping you organize, connect, and maintain a vault that would otherwise require significant ongoing manual effort.

This article is about practical workflows for using Claude with Obsidian as a second brain. It assumes you already have some familiarity with Obsidian basics. If you're starting from scratch on the plugins side, the first article in this series covers the best AI plugins and general AI workflows.

What "Second Brain" Actually Means Here

A second brain isn't just a note-taking app with a lot of notes in it. The meaningful version has a few properties:

Most people who try PKM do fine at capture and early organization. The system degrades when maintenance stops. Notes become orphaned. Categories become obsolete. The vault grows but becomes less useful.

This is where Claude is genuinely helpful. The maintenance work that people abandon is largely the kind of processing that AI is good at: finding patterns, suggesting connections, flagging stale material, reformatting raw captures into useful notes.

Vault Structure That Works with AI

Before getting into Claude workflows, it helps to have a vault structure that plays well with AI assistance. A few principles that the PKM community has converged on:

Keep notes in Markdown with clear titles. This sounds obvious but matters. Claude works with text. Notes with clear, descriptive titles and well-formed prose are much easier to work with than a pile of bullet fragments. Write notes the way you'd want to read them, not just as quick captures.

Use folders sparingly. Deep folder hierarchies are harder to work with. A flat or shallow structure with good tags and links works better for AI-assisted retrieval. Many serious Obsidian users use three folders: Inbox (unprocessed), Notes (processed), and Archive.

Separate capture from processing. Raw captures live in Inbox. Processed notes live in Notes. This distinction matters because you'll use Claude differently for each: rough processing for Inbox items, synthesis and maintenance for processed notes.

Consistent daily notes. If you're using daily notes (and most second brain practitioners do), keep a consistent format. Claude can work much better across a collection of daily notes when they follow a predictable structure.

Three Ways to Use Claude with Obsidian

There are three distinct integration levels, each with different trade-offs.

Option 1: Claude.ai or Claude in the Browser (No Setup)

The simplest approach: copy note content into Claude.ai and ask it to help. No plugins, no API keys, no setup.

This works better than most people expect. You can paste in a raw capture and ask Claude to turn it into a structured note. You can paste several related notes and ask for synthesis. You can paste a long journal entry and ask for the key themes.

What it's good for: One-off processing tasks where you have time to sit down and work. Works well for people with smaller vaults or who are just starting to experiment.

What it's bad for: Anything at scale. Copying and pasting manually doesn't work when you have a hundred Inbox items to process.

Option 2: Claude API via Obsidian Plugins

The Copilot for Obsidian plugin (covered in the first article) supports Claude directly. Add your Anthropic API key, set the default model to Claude, and you get a chat sidebar inside Obsidian with access to your current note or vault context.

This is the middle path. You get AI access inside your vault without writing any code. The trade-off is that you're working within the limits of what the plugin supports.

Best setup for this:

  1. Install Copilot for Obsidian via Community Plugins
  2. In settings, add your Anthropic API key and select Claude Sonnet 4 or Claude Opus 4 as your model
  3. Enable "Note as context" so every message includes your current note automatically
  4. Build a habit of opening the sidebar when processing a new note

What it's good for: Daily writing assistance, note refinement, single-note synthesis. The "chat about my note" workflow is genuinely useful.

What it's less good for: Batch processing, vault-wide operations, anything that requires working across many notes simultaneously.

Option 3: Claude Code for Vault-Level Operations

This is the power user path. Claude Code is a terminal-based AI coding agent that can read, write, and organize files. Since Obsidian vaults are just folders of Markdown files, Claude Code can perform vault-level operations that would be impossible through a plugin.

You run Claude Code in your vault directory and give it instructions like a capable assistant who can see all your files. Examples that actually work:

The practical setup:

  1. Install Claude Code: npm install -g @anthropic-ai/claude-code
  2. Set your API key: export ANTHROPIC_API_KEY=your-key-here
  3. Open your terminal, navigate to your vault directory
  4. Run claude and start asking

The key habit is treating Claude Code as a vault assistant, not a coding tool. It happens to be built for software development but works equally well for text file operations.

Daily Workflows That Actually Stick

The second brain workflows that work are the ones that fit into existing habits. Here are patterns that Obsidian users report using consistently.

Morning Inbox Clear

The problem: You have 15 items in your Inbox captured yesterday. Processing them manually takes 20-30 minutes. You skip it.

The workflow with Claude:

  1. Open Claude Code in your vault directory
  2. Ask: "Read through my Inbox notes and for each one, suggest a title, a one-line summary, relevant tags, and any notes it should link to based on my existing Notes folder. Output your suggestions in a format I can review."
  3. Review the suggestions, edit as needed, then act on the ones that make sense.

This doesn't eliminate your judgment. You're still deciding what goes where. But the initial processing is done, and you're editing rather than creating from scratch. That's a meaningful difference in friction.

Weekly Synthesis

The problem: You've captured a lot this week but haven't extracted the insights. The raw notes exist but the synthesis doesn't.

The workflow with Claude:

  1. Give Claude Code access to your daily notes from the past week
  2. Ask: "What are the three to five main themes in my notes this week? For each theme, what are the key notes I should connect or consolidate?"
  3. Use that output as the framework for your weekly review, filling in your own reflection.

The Claude output gives you a starting structure. You validate, revise, and add the personal judgment it can't provide.

Stale Note Audit

The problem: Your vault has notes you haven't touched in six months that probably need updating or deleting.

The workflow with Claude:

  1. "Find notes that haven't been linked to any other note and haven't been modified in more than 90 days. List them by folder, with a one-line summary of what each one is about."
  2. Review the list and make decisions: archive it, update it, link it, or delete it.

This turns a vague maintenance task into a concrete review list.

Meeting and Research Processing

The problem: You have rough notes from a meeting or a reading session. They're too raw to be useful but need work to become proper notes.

The workflow with Claude: Open the raw note in Obsidian, activate the Copilot sidebar (with your Claude model), and ask:

The constraint "don't add anything that isn't in the raw notes" is important. Without it, Claude will sometimes embellish with plausible-sounding details that aren't accurate.

Building a Map of Content with Claude

Maps of Content (MOCs) are notes that act as hubs: they don't contain atomic ideas themselves, but link to a cluster of related notes and provide an overview. They're one of the most useful structures in a second brain, and also one of the most tedious to build manually.

Claude Code handles this well:

`` "Create a Map of Content note for my notes about machine learning. Review all notes tagged #ml or #machine-learning, identify the main subtopics, and write a MOC note with sections for each subtopic linking to the relevant notes. Use the actual note titles for links." ``

The output won't be perfect. Claude may group things differently than you would. But having a draft MOC that you then edit is significantly faster than starting from scratch.

What to Watch Out For

Claude inventing details. When processing notes, Claude will occasionally add context or details that sound plausible but aren't in your original notes. Always review processed notes against your source material, especially for factual content.

Over-organization. AI makes organization effortless, which can lead to over-engineering the system. A vault with 40 tags and 200 MOC notes is not necessarily better than one with 10 tags and 20 MOC notes. The goal is useful retrieval, not perfect taxonomy.

Losing your voice. Notes that have been heavily rewritten by AI start to sound the same. For personal notes, journaling, and thinking-in-progress, do the writing yourself. Use Claude for processing and synthesis, not for generating first-person content about your own thoughts.

Trusting Claude's memory of your vault. Claude's context window is limited. In a very large vault, Claude Code can't read every file at once. For vault-wide operations, be specific about which folders or files to look at.

Practical Starting Point

If you're new to combining Claude with Obsidian, the path of least resistance:

  1. Start with Claude.ai in the browser. Pick five notes from your Inbox and paste them in one at a time. Ask Claude to turn each into a well-structured note with a title, summary, and relevant tags. Get a feel for the quality before investing in plugins.
  1. If you like the results, install Copilot for Obsidian and add an Anthropic API key. Use it for single-note work.
  1. Once you have more than 200 notes and feel the maintenance burden, try Claude Code for one vault-level task: the stale note audit or the weekly synthesis. That will show you whether the power user path is worth it for your system.

There's no correct answer on which level to use. Some people with 5,000 notes are happy with the browser-based approach. Some people with 200 notes want Claude Code. It depends on how seriously you're managing your vault and how much you value automation versus simplicity.

This is the second article in the AI + Obsidian series on Solaire Tools. The first article covers the best AI plugins for Obsidian, including Smart Connections, Copilot for Obsidian, and Text Generator.