How to Build a Second Brain ChatGPT Can Actually Use

You run half your working life through ChatGPT. Drafts, plans, decisions, research. And every single conversation starts from zero.

So you re-explain. Who you are, what you sell, who your customers are, what you decided last month. The same paragraph, typed again, slightly worse each time.

A second brain fixes that. Not the built-in memory feature. An actual knowledge base that ChatGPT can search, read, and write.

What a second brain for ChatGPT is

A second brain is a place outside your head where you keep everything worth keeping: notes, research, decisions, things you learned.

A second brain for ChatGPT adds one requirement: ChatGPT has to be able to read it. Not through copy-paste. Not through file uploads you redo every week. Through a live connection where ChatGPT searches your notes mid-conversation and pulls in exactly what it needs.

That connection is MCP, the Model Context Protocol. It's an open standard for letting AI assistants use external tools. Your knowledge base runs an MCP server; ChatGPT connects to it; your notes become something ChatGPT can look up the way it looks up the web.

Why ChatGPT's built-in memory isn't one

ChatGPT already has a memory feature. It collects short facts as you chat: your name, your preferences, fragments of projects.

Useful, but it is not a second brain, for three reasons:

1. It holds crumbs, not documents. "User runs a design studio" is a fact. Your positioning doc, your pricing rationale, and twelve customer interviews are knowledge. Built-in memory was never designed to hold the second kind.
2. You can't really work with it. You can view and delete memories, but you can't organize them into folders, version them, or edit them like documents. When something changes, you hope the old fact fades.
3. It's locked to one app. Your ChatGPT memory means nothing to Claude, Perplexity, or the coding assistant in your editor. A second brain should follow you.

A knowledge base flips all three: real Markdown notes you own, organized in folders you chose, readable by every assistant you connect. See the full Hjarni vs ChatGPT memory comparison.

What to put in it

Don't start by importing everything you've ever written. Start with the five notes you keep re-explaining:

• Who you are and what you sell. Your one-paragraph pitch, your offer, your prices.
• Who your customers are. Your ICP, plus the phrases real customers use. Paste in interview quotes.
• Decisions and why you made them. Pricing changes, positioning calls, things you tried that failed. This is the note that stops ChatGPT from re-litigating settled questions.
• How you sound. Your style guide: words you use, words you ban, three example paragraphs in your voice.
• The current state of things. What you're working on this month. Update it as you go, or let ChatGPT update it for you.

Writers keep style guides and research per project. Founders keep strategy, investor notes, and meeting summaries. The pattern is the same: write down what you keep repeating.

How to set it up

Five minutes, no code:

1. Create a knowledge base. Sign up for Hjarni (free to start, no credit card) and write those first five notes in plain Markdown.
2. Connect ChatGPT. Install Hjarni from the ChatGPT Apps Directory: one click, then sign in. The full setup guide covers the developer-mode path too.
3. Test it. New conversation, then: "Search my notes for our positioning and rewrite this email to match it." Watch it pull the actual note.
4. Close the loop. End your next work session with: "Save what we decided as a note in my Decisions folder." ChatGPT writes it, with your approval, and the next conversation starts already briefed.

What it looks like day to day

The difference shows up in small ways, constantly.

You ask for a landing page draft and it matches your positioning without being told. You ask "what did we decide about annual pricing?" and get your own reasoning back, not a generic pros-and-cons list. You start Monday with "read my current-state note and draft this week's plan" and it works from what you actually wrote.

Your context compounds instead of resetting. That's the whole trick.

And yes, it works with Claude too

Because MCP is an open standard, the same second brain serves every assistant. Connect Claude and it reads the same notes ChatGPT wrote. Switch tools next year and your knowledge comes with you. That's the MCP second brain pattern: one brain, every assistant, no lock-in.

ChatGPT is where most people live. Your knowledge shouldn't be trapped there, or anywhere.

---

Write once. You both remember. Start your second brain.