Building a Personalized Second Brain with Claude Code and Obsidian

Have you ever taken notes on an issue, only to have to rethink it from scratch a few weeks later because you couldn’t find them? Or opened Notion to find hundreds of pages that no one ever reads again? That’s a sign of a passive “second brain”—it merely holds information without doing anything with it.

This article guides you through building a completely different system: combining Obsidian (a local Markdown-based note-taking app) with Claude Code (an AI agent running directly in your terminal, reading and writing files on your machine) to create an actionable digital brain—one that autonomously maintains context across work sessions, automatically processes your Inbox, and synthesizes information instantly whenever you need it.

Goal of this article: After completing this guide, you will possess a practical, operational system—from the Vault folder structure and a standardized CLAUDE.md configuration file to 4 automated workflows tailored for your daily work.

1. Why Traditional Note-Taking Systems Fall Short

Most “second brains” people build suffer from a common weakness: passivity. You jot things down, but the data just sits there. It doesn’t connect itself. It doesn’t remind you when needed. And more importantly—it doesn’t proactively help you take action.

According to Tiago Forte’s PARA method, an ideal second brain should store and retrieve information precisely when you need it most. However, even powerful tools like Obsidian remain passive storage repositories without an underlying artificial intelligence to synthesize and process the data.

2. How is an AI brain different from a passive note repository?

While traditional note-taking tools serve merely as static storage, the AI brain combining Obsidian and Claude Code delivers four superior, proactive capabilities:

• Multidimensional synthesis and linking: Automatically reads data from multiple fragmented notes, detects commonalities, and connects them into a comprehensive big picture.
• Learning personal habits: Recognizes and remembers your working style across sessions instead of completely forgetting it after every chat.
• Proactive action: Drafts content, summarizes documents, categorizes tasks, and plans schedules based directly on the data repository you have saved.
• Maintaining continuous history: Automatically logs work after each session so that the next session can pick up the train of thought immediately.

3. The Core Tool Duo: Obsidian and Claude Code

Obsidian – Local and Secure Data Storage

Obsidian is a note-taking application that stores all data as Markdown files (.md) directly on your local hard drive—without relying on third-party cloud servers. This means you have full control and 100% ownership of your data. The key takeaway is: because the data resides right on your computer, Claude Code can access, read, and update files in your vault directly and securely.

Claude Code – Command-Line AI Agent

Claude Code is a command-line AI tool from Anthropic that runs directly in the terminal—completely different from the traditional Claude Web version. When running Claude Code directly inside an Obsidian vault folder, the agent gains exceptional capabilities:

• Access and read the entire folder tree and note files within the vault.
• Write new files, update the content of existing files, and automatically clean up directories.
• Execute shell commands, verify logic, and batch-process documents.
• Retain context through locally stored log files.

3 preparation steps before getting started

Before diving into setting up the system, you need to prepare the following 3 elements:

1. Install the Obsidian application and initialize a new note Vault on your computer.
2. Install the Claude Code CLI in your terminal and start a session.
3. Set aside about 30–60 minutes to standardize the Vault folder structure and write the CLAUDE.md instruction file.

4. Standardizing Obsidian Vault Structure and Frontmatter

Claude Code works most effectively when your vault is clearly structured, consistent, and predictable. Below is a directory structure based on an upgraded PARA method (adding an /AI folder):

/MyVault
├── /Inbox          ← Draft notes, unprocessed quick captures
├── /Projects       ← Dedicated folders for active projects
├── /Areas          ← Ongoing responsibilities (health, finance, work)
├── /Resources      ← Reference materials organized by topic
├── /Archive        ← Completed projects, archived notes
├── /AI             ← AI outputs: session logs, drafts, summaries
└── /Templates      ← Reusable note templates

The /AI Directory: The Key to Maintaining Workflow Continuity

The /AI directory is a particularly important component that most standard vaults lack. This is where Claude Code automatically stores output after each session—including session summaries (session logs), document drafts, and work logs. Over time, this directory becomes a chronicle recording everything the AI has done for you.

Standardizing Frontmatter (Metadata Info Block)

Frontmatter is the metadata block placed at the top of each note file, enclosed between two lines of hyphens ---. It acts as an instructional tag that helps Claude Code quickly categorize notes without having to read the full content of hundreds of files:

---
created: 2026-10-15
tags: [project, devpixel, active]
status: in-progress
related: ["[[Hop-kick-off-thang-10]]", "[[Brief-du-an-DevPixel]]"]
---

The related field utilizes Obsidian’s wiki-link syntax ([[Note Name]]). Claude Code can parse these links to construct an interconnected contextual map across documents before executing any action.

4 Recommended Obsidian Community Plugins

For the system to run smoothly, you should install the following 4 community plugins:

• Templater: Automatically creates Daily Notes and standardized note templates.
• Periodic Notes: Manages daily, weekly, and monthly note cycles with consistent naming conventions.
• Dataview: Allows you to query data within your vault like a structured database.
• QuickAdd: Quickly captures ideas into your Inbox from anywhere using hotkeys.

5. Setting Up the CLAUDE.md File: The Operating Heart of the AI

The CLAUDE.md file is located right in the root directory of your Vault. This is a permanent instruction document that Claude Code automatically loads every time you open a new session in the terminal. If you invest the effort to write this file well, every subsequent response and action from the AI will be extremely accurate.

4 Core Components in the CLAUDE.md File

A standard CLAUDE.md file should include 4 main sections:

1. About Me: Describe your role, responsibilities, and work goals so the AI understands your professional perspective.
2. Vault Structure: Explicitly outline folder locations so the AI knows where to find information and where to save outputs.
3. Session Protocols: Clearly define actions for starting a session (reading the Daily Note, checking the Inbox) and ending a session (saving session logs, updating the Daily Note).
4. Output Preferences: Specify the language (Vietnamese), presentation style (prioritizing bullet points), and the confirmation protocol required before modifying files.

# AI Second Brain - Agent Instructions

## About Me
I am a specialist at DevPixel. I manage 3-4 parallel projects and need assistance synthesizing research, drafting technical documentation, and tracking key decisions.

## Vault Structure
- /Inbox: Unprocessed notes - check here first
- /Projects: Each project has its own folder; active projects have status: active
- /AI/sessions: Log session summaries after each working session

## Session Protocols
When starting:
1. Read today's Daily Note (format: /Inbox/YYYY-MM-DD.md)
2. Scan /Inbox to see if there are any unprocessed notes
3. Check notes tagged with #needs-review

When ending:
1. Write the session summary to /AI/sessions/YYYY-MM-DD-HH.md
2. Update the "AI Log" section in today's Daily Note

## Output Preferences
- Write in Vietnamese: clear, concise, avoiding unnecessary technical jargon
- Prefer short bullet points over long paragraphs
- If information is uncertain: ask for confirmation before modifying files

Core Takeaway: An overly vague CLAUDE.md = poor AI output. Generic instructions like “Help me do my work” are completely useless. The quality of your AI second brain is directly proportional to the level of detail and specificity in this file. Whenever you notice the AI not responding as expected, immediately update CLAUDE.md with a new rule.

6. Maintaining Long-Term Memory via Session Logs

Claude Code itself does not have a built-in long-term memory mechanism between terminal restarts. System continuity relies entirely on Session Logs saved directly into the Vault.

At the end of each session, Claude Code automatically writes a summary file to /AI/sessions/YYYY-MM-DD-HH.md containing:

• Key tasks completed and decisions made during the session.
• Open loops or issues that need further follow-up.
• Current status of each related project.

When you start a new session the next day, the AI only needs to skim the most recent session log to immediately understand what you were working on yesterday, which projects are on hold, and your formatting preferences.

The 2-Minute Habit: Before starting a new session, spend 2 minutes reviewing the previous session log. This helps you quickly catch any AI reasoning errors and fine-tune your CLAUDE.md file before those mistakes turn into entrenched habits.

7. 4 Practical Workflows to Optimize Your Work

Below are the 4 automation workflows that deliver the most immediate results, which you can apply right away to your data warehouse:

Workflow 1: Daily Inbox Processing (Inbox Zero in 5 Minutes)

Every morning, all you need to do is dump every raw note into the /Inbox folder—from transcribed voice memos and important emails to whiteboard photos and quick notes on your phone. Then open your terminal and type a single simple command:

claude "Process all new notes in the Inbox folder."

The AI will read through each note, automatically generate fully structured files in the appropriate directories (/Projects or /Resources), append complete frontmatter, tags, and wiki-links, and then archive or delete the original drafts. A task that previously took 20–30 minutes is now completed in less than 5 minutes.

Workflow 2: Automating Weekly Reviews

The Weekly Review is the most valuable habit, yet it’s also the easiest to abandon because compiling everything takes so much time. With Claude Code, all you need to do is run a single prompt:

claude "Summarize this week's Weekly Review based on Daily Notes and Session Logs."

The AI will scan through all daily notes from the week, extract the list of completed tasks and pending issues, and draft a well-structured weekly report. You only need to spend 5 minutes skimming, making adjustments, and confirming the plan for the upcoming week.

Workflow 3: Pre-Meeting Preparation (Meeting Brief)

Before an important meeting with a partner or client, trigger the command:

claude "Prepare a meeting brief for the 2:00 PM meeting with [Client / Partner Name]."

Claude Code will automatically search across the entire vault for all notes related to that partner, extract project progress, and summarize commitments as well as open questions from previous meetings. You will walk into the meeting room thoroughly prepared without having to spend time searching through individual files.

Workflow 4: In-Depth Research and Literature Synthesis

When starting a new project or researching an in-depth topic, you simply need to feed Claude Code a set of reference materials (PDF files, markdown articles, web links):

claude "Research the documents in /Resources/Research, extract core arguments, and link them to active projects."

The AI will read through all the documents, extract key takeaways, identify patterns, and automatically generate linked notes in /Resources, ready to serve as input for subsequent working sessions.

8. Operational Principles and Data Security

To ensure your Second Brain system operates stably and securely over the long term, keep the following 5 important principles in mind:

• Start small: Don’t try to implement all workflows at once. Choose the single process you currently find most time-consuming (usually Inbox processing) and let the system run smoothly for 1–2 weeks before expanding.
• Maintain Vault consistency: Inconsistency in naming conventions or storage structures creates significant friction for AI. Spending an hour standardizing your vault initially will save you hundreds of operational hours down the line.
• Continuously update CLAUDE.md: Treat this file as a living document. Whenever the AI misunderstands a request, update the file immediately with specific rules so it doesn’t repeat the mistake.
• Don’t overuse AI for trivial tasks: If you only need to fix a typo or rename a file, do it manually. Claude Code delivers the most value when synthesizing multiple notes, handling multi-step workflows, or drafting complex content.
• Security and privacy: Obsidian stores files locally, so your data is not automatically uploaded to the cloud. However, the contents of your prompts and the files you instruct the AI to read will be sent to Anthropic’s API for processing. For highly sensitive documents (legal contracts, passwords, confidential financial data), carefully consider whether to include them within the AI’s scanning scope.

9. Frequently Asked Questions (FAQ)