What is a second brain for AI?

A second brain for AI is a structured personal knowledge base — typically in Obsidian — that stores your project context, decisions, notes, and lessons outside of any AI chat window. You give the AI access to this vault at the start of each session so it already knows your background, saving you from re-explaining your project every time you open a new chat.

Why use Obsidian with Codex or Claude Code?

Obsidian stores your project context in plain markdown files that Claude Code and Codex can read directly. By giving the AI your vault path via the starter prompt, every session starts informed — no repeated context, no wasted tokens re-explaining your project from scratch. The AI knows your decisions, your clients, your architecture, and your previous lessons.

Does Obsidian replace ChatGPT memory or ChatGPT Projects?

No — different scope. ChatGPT memory and Projects store context inside OpenAI's system. An Obsidian vault is richer (1,000+ notes, linked structure, full markdown documents), completely portable, and not locked to any single AI model. You can use the same vault with Claude Code, Codex, Gemini, or any future model.

Do I need Replit and GitHub from day one?

No. Start with just Obsidian + Claude Code. Build your vault and get comfortable with the prompt workflow first. Add Replit when you are ready to build and run code, and add GitHub when you are shipping — or when you want an audit trail of every change before it goes live.

Where should I store my Obsidian vault for safety?

Store your vault in a synced folder: Google Drive, iCloud Drive, or Dropbox. This keeps it automatically backed up and accessible across devices. Never store the vault only on your local disk — a single drive failure means losing everything you have built in your second brain.

Can I store passwords or API keys in my Obsidian vault?

No. Your Obsidian vault is plain-text markdown files stored on your disk and synced to cloud storage. Never paste passwords, API keys, tokens, or other secrets into vault notes. Use a dedicated password manager (1Password, Bitwarden, or similar) for all credentials.

How does this reduce repeated AI context and wasted tokens?

The starter prompt loads your vault context at the start of every session. The AI reads your project background, previous decisions, and relevant notes before you ask your first question. You skip the 5–10 minute re-introduction every founder wastes on every new AI session — and you stop paying for the tokens that re-introduction costs.

What is the difference between Obsidian, Codex, Replit, and GitHub in this workflow?

Obsidian is your memory layer — all your project context, decisions, and notes live here in plain markdown. Claude Code or Codex is your engineering intelligence layer — it reads the vault and plans or writes code informed by your context. Replit is your build and runtime layer — code runs and deploys here. GitHub is your audit and review layer — every commit is a checkpoint, and humans review changes before anything ships.

Obsidian Second Brain for Codex, Replit & GitHub

TL;DR

The four-layer AI workflow gives you persistent memory (Obsidian vault), engineering intelligence (Claude Code or Codex reads the vault before every session), a build runtime (Replit executes tasks), and an audit trail (GitHub commits every change). Drop the starter prompt at the start of each session, let the AI read your vault, and stop explaining the same context from scratch every time you start a new chat.

In this article

01Why every new AI chat wastes your time
02The four-layer AI workflow
03Step-by-step: setting up Obsidian
04The AI Second Brain Prompt Pack
05See it in action — Nakul's live vitiv.ai vault
06Download the tools
07Want this workflow set up for your team?

If your AI has no memory, it keeps starting from zero. If your AI has no receipts, you cannot trust what changed. This is the workflow that fixes both.

TL;DR

Start with Obsidian and Codex (or Claude Code) — that is all you need right now. Zero repeated context, no re-introduction every session. Replit and GitHub come in when you are ready to build and ship code. Two tools today, four tools when you need them.

Why every new AI chat wastes your time

Every time you open a new chat window, you start from scratch. The AI does not know your project, your decisions, or what you tried last week. Even the most powerful AI agents and automation tools only know what you tell them in each session. Here is what that costs you:

New chat = zero context. You explain your project from the top every single session.
Long threads become impossible to follow as the conversation grows.
Context windows fill up and earlier decisions silently drop out.
You repeat the same background for every new session, every new tool.
Token costs rise because you are re-sending the same information repeatedly.
Mistakes from last week get repeated this week because the AI has no memory of them.

The four-layer AI workflow

Each tool does one job. Start with the first two — Obsidian and Codex give you persistent memory and engineering intelligence today. Replit and GitHub join when you are ready to build and ship: they add a live runtime and an audit trail. Same pattern behind every AI automation project that actually scales.

Obsidian

Memory layer

Your vault. Every decision, note, prompt, and lesson lives here — outside any chat window.

Codex (OpenAI)

Engineering & planning layer

Codex reads your vault before every session. No repeated context. It already knows your project. (Claude Code works the same way — same pattern, different model.)

Replit

Build layer — when you start building

When you are ready to build: code runs here. Replit agents execute tasks that Codex or Claude planned against your vault context.

GitHub

Review layer — when you start shipping

When you are ready to ship: every change is committed. GitHub is your audit trail — humans review before anything goes live.

Start with Obsidian + Codex. Add Replit + GitHub when you are ready to build.

Step-by-step: setting up Obsidian

Obsidian app interface — free, local-first note editor for Mac, Windows, and Linux

Obsidian: free, local-first, and available on every platform.

Download and install Obsidian (free)

Obsidian is free, local-first, and available for Mac, Windows, and Linux. Download Obsidian

Create a new vault

Open Obsidian → "Create new vault". Give it a name like "MySecondBrain" or "WorkVault".

Store it in a synced folder

Place the vault inside Google Drive, iCloud Drive, or Dropbox so it stays backed up. Avoid storing it only on a local disk.

Copy the vault path

Mac example: /Users/yourname/Google Drive/MyVault Windows example: C:\Users\yourname\Google Drive\MyVault

Paste the path into Codex or Claude Code

Open Codex at platform.openai.com/codex (or Claude Code). Use the prompt pack below and replace [PASTE YOUR OBSIDIAN VAULT PATH HERE] with your real path.

Mac Finder showing technokul_vault inside Google Drive

The vault folder lives inside Google Drive — always synced, never only on a local disk.

The AI Second Brain Prompt Pack

Do not use only one tiny starter prompt. Use this as a system: first set up the vault, then store information properly, then start every new AI session by reading the vault.

Create vault

Download Obsidian and create a new vault in a synced folder (Google Drive, iCloud, or Dropbox). Your AI cannot remember anything it cannot reach.

Run setup prompt

Paste Prompt 1 into Codex or Claude Code. It builds the folder structure and root notes — non-destructively, checking for existing files first.

Store information properly

At the end of any useful session, use Prompt 2 to save decisions, notes, and tasks back into the correct vault note.

Use the new-chat starter every session

Paste Prompt 3 at the start of every new Codex or Claude Code chat. The AI reads your vault before asking questions.

Prompt 1 — First-Time Vault Setup

I want to use Obsidian as my second brain so AI tools do not start from zero every time.

My Obsidian vault path is:
[PASTE YOUR OBSIDIAN VAULT PATH HERE]

Your job:
Set up a practical AI-readable Obsidian vault structure for my work, projects, notes, decisions, prompts, and reusable context.

Before making changes:
1. Inspect the vault path.
2. Tell me whether the vault is empty or already has files.
3. If files already exist, do not overwrite them.
4. Ask before changing any existing note.
5. Do not store passwords, API keys, private keys, tokens, cookies, recovery codes, or sensitive credentials.

Create this starter folder structure if it does not already exist:

- 00 Inbox
- 01 Indexes
- 02 Daily Notes
- 03 Projects
- 04 Clients
- 05 Companies & Ventures
- 06 Systems & Infrastructure
- 07 Assets & Hardware
- 08 Finance & Legal
- 09 Content & Marketing
- 10 Personal Systems
- 11 Knowledge Base
- 12 Templates
- 13 Archive
- 99 Attachments

Create these root notes if they do not already exist:

- Home.md
- Start Here.md
- README_FOR_AI.md
- Where To Put Things.md
- Current Active Projects.md
- Master Task List.md
- Decision Log.md
- Prompt Library.md

Work carefully:
1. First show me your plan.
2. Then create only missing folders/files.
3. Do not overwrite existing notes.
4. After creating the structure, summarize what you created and what I should fill in first.

⚠Do not paste passwords, API keys, or tokens into your vault or this prompt.

Prompt 2 — Store This Properly

Use my Obsidian vault as the source of truth.

Vault path:
[PASTE YOUR OBSIDIAN VAULT PATH HERE]

Before storing anything:
1. Read README_FOR_AI.md if it exists.
2. Read Where To Put Things.md if it exists.
3. Search the vault for existing notes related to this topic.
4. Do not create duplicates if a relevant note already exists.
5. Do not store passwords, API keys, private keys, tokens, cookies, login codes, or sensitive credentials.

Information to store:
[PASTE THE INFORMATION, SUMMARY, TRANSCRIPT, DECISION, IDEA, OR TASK HERE]

What I want you to do:
1. Identify what type of information this is.
2. Decide where it belongs.
3. Search for existing relevant notes.
4. Tell me what you found.
5. Append to a suitable note or propose a new note path.
6. Use Obsidian links.
7. If this contains a decision, update Decision Log.md.
8. If this creates tasks, add them to the right project note or Master Task List.md.
9. Ask only missing questions if unclear.
10. Tell me what changed and why.

⚠Do not paste passwords, API keys, or tokens into your vault or this prompt.

Prompt 3 — New Chat Starter

Before helping with this task, use my Obsidian vault as the source of context.

Vault path:
[PASTE YOUR OBSIDIAN VAULT PATH HERE]

Start by reading:
- README_FOR_AI.md
- Home.md
- Start Here.md
- Where To Put Things.md
- Current Active Projects.md
- Decision Log.md
- Prompt Library.md

Then search the vault for notes relevant to my request.

Workflow:
1. Understand my task.
2. Search the vault before asking me questions.
3. Summarize what you found.
4. Surface recorded decisions that may affect this task.
5. Flag conflicts between my request and the vault.
6. If context is missing, ask concise questions.
7. If the task is clear, work in phases:
   - understand and plan
   - implement or draft
   - verify
   - document
8. After useful work, update or suggest updates to the correct Obsidian note.
9. Do not store secrets or credentials.
10. Do not delete or overwrite notes unless I explicitly ask.

For this session, my request is:
[WRITE YOUR CURRENT TASK HERE]

⚠Do not paste passwords, API keys, or tokens into your vault or this prompt.

Prompt 4 — Project Build Planning

Use my Obsidian vault as context before planning this build.

Vault path:
[PASTE YOUR OBSIDIAN VAULT PATH HERE]

Project:
[PROJECT NAME]

Goal:
[WHAT I WANT TO BUILD OR CHANGE]

First:
1. Read README_FOR_AI.md and Where To Put Things.md.
2. Search for the project note.
3. Find existing decisions, tasks, bugs, prompts, and links.
4. Tell me what you found.

Then create a phased plan:
1. What already exists
2. What needs to be clarified
3. What should be built first
4. What should not be touched
5. What files/tools/repos are likely involved
6. What needs manual review
7. What should be documented back into Obsidian

Do not start implementation until I approve the plan.

⚠Do not paste passwords, API keys, or tokens into your vault or this prompt.

Need help setting this up for your team?

We set up Obsidian + Codex workflows for founders, consultants, and small teams. Talk to vitiv.ai on WhatsApp — no pitch, just a practical conversation about your setup.

Codex reading the Obsidian vault context — the prompt pack in action

Codex (or Claude Code) reads vault orientation files before answering — vault context loads in seconds, zero repeated explanation.

See it in action — Nakul's live vitiv.ai vault

This 15-second clip zooms slowly into the Obsidian graph view of vitiv.ai's own second brain — 1,000+ interconnected notes on clients, projects, decisions, and AI workflows.

▶ Click to play

vitiv.ai Obsidian vault — second brain demo — click to play

Obsidian graph view — vitiv.ai second brain with 1,000+ interconnected notes

The vitiv.ai second brain: 1,000+ notes across 19 active projects, linked by concept rather than folder.

Download the tools

Obsidian — the vault (free)

obsidian.md — local-first, markdown, free on Mac, Windows, and Linux. The best AI-ready note-taking app for this workflow.

Codex (OpenAI) — primary AI planner

platform.openai.com/codex — reads your vault, plans work, writes code. Ideal for custom AI tools and project builds.

Claude Code — the alternative

claude.ai/download — same pattern as Codex, Anthropic model stack. Use whichever you prefer.

Replit — the build runtime (optional)

Add when you start building. replit.com — run, deploy, and iterate without leaving the browser.

GitHub — the audit trail (optional)

Add when you start shipping. desktop.github.com — every commit is a checkpoint. Humans review before anything goes live.

Want this workflow set up for your team?

Talk to vitiv.ai on WhatsApp

We set up Obsidian + Codex + Replit workflows for founders, consultants, and small teams. WhatsApp us to discuss your setup — no pitch, just a practical conversation.

Explore vitiv.ai services: AI automation · AI agents · Custom AI tools · AI-ready websites · Back to blog

Free strategy call

Thinking about your first AI deployment?

30 minutes. We sketch the architecture, cost, and timeline — at no charge.

Talk to us

How I Use Obsidian as a Second Brain for Codex, Replit, and GitHub

Why every new AI chat wastes your time

The four-layer AI workflow

Step-by-step: setting up Obsidian

The AI Second Brain Prompt Pack

See it in action — Nakul's live vitiv.ai vault

Download the tools

Want this workflow set up for your team?

FAQs

Want this in your business?

Generative Engine Optimization (GEO) in 2026: The Complete Guide to Ranking in ChatGPT, Perplexity & Google AI Overviews