A second brain is supposed to do the remembering for you: whatever you want to keep goes into one place, and your head stays free for actual work. In practice it often turns out differently. The notes are there, hundreds of them, yet when you need something you still spend ten minutes looking for it.

What a second brain is

The term was popularized mainly by Tiago Forte, first through his online course and then, in 2022, through his book “Building a Second Brain”. The premise: your brain is for thinking, not for storage. Anything worth keeping belongs in a digital system you trust, so you don’t have to carry it around in your head.

Forte describes the workflow in four steps, abbreviated CODE:

  1. Capture: keep what strikes you as useful, rather than hoarding everything.
  2. Organize: file things by what you’ll use them for, not by subject.
  3. Distill: pull out the essence so you can grasp it quickly later.
  4. Express: turn what you collected into your own work.

For filing he proposes PARA: Projects (efforts with a goal and an end), Areas (responsibilities with no end date), Resources (topics you’re interested in) and Archives (everything finished or inactive). Most are built in Notion, Obsidian, Evernote or similar apps.

The problem: an archive doesn’t answer

Collecting usually goes fine. Retrieval is where it breaks down. A classic second brain is passive: it stores things but does nothing on its own. When you have a question, you need to know that a note on it exists, roughly remember where it lives, and guess the right search term. Full-text search finds “notice period” but misses the note that talks about “contract term”.

Then there’s upkeep. Distilling, meaning summarizing and highlighting, is manual work, and manual work is the first thing to slip in a busy week. Documents are often a blind spot too: a scanned contract sits in a folder as an image, and the only searchable thing about it is the file name.

What changes when your AI app can read along

Through MCP, an open standard, AI apps like Claude or ChatGPT can reach external data sources. A service exposes its knowledge once as an MCP server, and every AI app that speaks the protocol can search it.

For a second brain, that means you stop searching and start asking. “What did we agree with the landlord about the service charge statement?” The AI app searches your notes and documents, reads the relevant passages and answers. The last part matters most: a useful answer gives its citations, meaning which note or document, ideally with the page. That way you can check instead of trusting the AI blindly.

The distill step moves as well. The AI app can write summaries, link notes, or file what came out of a conversation straight away. What stays with you is deciding what goes in. A second brain full of random clippings gives random answers, with or without AI.

Classic second brainWith an AI app over MCP
Finding thingsYou search for keywordsYou ask a question
SummarizingBy handThe AI app drafts, you review
Scans and PDFsOften just the file nameSearchable if the service runs OCR
ResultA list of hitsAn answer with citations

Where Memexa fits

Memexa is one way to set this up, not the only one. Notes, documents and scans become a linked knowledge graph, and scans go through OCR. Claude, ChatGPT and any other AI app with MCP read the memory via https://brain.memexa.ai/mcp, and answers come with citations down to the page. Hosting is in the EU, and you can export everything as Markdown at any time. Plans start at €2.49 a month billed yearly or €2.99 monthly (pricing); there is no free plan, and the 14-day trial requires a card.

If your second brain is an Obsidian vault of plain text notes on a single computer, you don’t need a hosted service: Claude can read the folder directly. The options are laid out in Connecting Obsidian and Claude. A hosted memory pays off mainly when a lot of documents are involved, or when you want to reach it from several devices and AI apps. What that looks like with insurance, warranties and school letters is in Find your paperwork with AI.

How to start

  1. Start with one area, not everything. A current project or the paperwork for your flat is enough.
  2. Store what you’ll later ask about: agreements, decisions, contracts, sources.
  3. Use a light version of PARA: one folder or tag per active project, finished things into the archive.
  4. Connect your AI app via MCP: enter the address, sign in once in the browser. Claude, ChatGPT and Mistral Vibe allow a custom connector even on the free plan (Claude exactly one). The individual steps are on How it works.
  5. Test with questions you already know the answer to. If the citations hold up, you can trust the system with the questions you don’t know the answer to.

Frequently asked questions

Can I keep my existing second brain?

Yes. An Obsidian vault imports into Memexa as a ZIP with its wikilinks intact, and Notion exports Markdown. Many people keep writing in their usual tool and use the memory as a layer their AI app reads.