What is an AI second brain?

A place where what you read, write and decide is kept and can be asked. Here it is a set of vaults on your own computer: notes, documents, web pages and conversations, indexed locally, answered by the model you choose, with the source named every time.

Does it work offline?

Yes. Indexing and search run on your machine, and with a local model downloaded once, the answers do too. A cloud model is a choice you make per conversation.

Does it replace Obsidian or Notion?

No. It connects to them. An Obsidian or Logseq vault is a folder of Markdown files you give to a vault to watch; a Notion export lands the same way. It reads your notes and never writes into them.

Where are my notes stored?

In a folder you choose, as plain Markdown files, with wiki links and headings. Any editor opens them. The memory index is another file on your disk. Nothing is uploaded.

Does it remember my conversations?

Yes, by default: each conversation becomes one memory in a Conversations vault, kept current as it grows. You can switch it off per conversation, and deleting a conversation asks whether to forget its memory too.

Are my notes encrypted?

They can be. AES-256 encryption at rest exists and you arm it in the settings. By default your vaults are local files in the clear, readable without us.

Second brain

An AI second brain. Local, and yours.

Notes, documents, web pages and conversations, filed in vaults, one per domain of your life, on your own computer. You ask, it answers from your own notes and documents, and names the source. No account. No cloud required.

Download, free How it works ↓

The problem

Change the model. Keep the brain.

You spent years filing. The tool changed its pricing, its export, or its mind. The notes are still yours in theory; in practice they live where you cannot ask them anything without paying, and where a model you did not choose reads them.

Mnemosyne OS keeps the memory on your machine, in files you can open without us, and lets you plug any model into it. Privacy-first knowledge management, open core: the SDK and the CLI are open, the memory engine is sealed.

A second brain on someone else's server is a subscription with your memory in it.

Mnemosyne OS · mnemosyne-os.io/second-brain Quote this line freely, with the link.

One brain, several rooms

Your work does not read your journal. Unless you say so.

A vault is one domain: work, research, health, the family. Each is its own file. A question reads them all, or the one you name. A vault marked MAXIMUM answers only when named, and its content is not sent to a cloud model unless you confirm it, each time.

An attention dial per vault says how loudly it answers. The Neural Map draws what you have accumulated as a graph: where the clusters are, where two domains touch, where nothing is. And a nightly pass links related memories across vaults, so a note from 2023 can resurface next to a document from last week.

Privacy-first notes

Your notes are files on your disk. The AI reads them there.

The built-in editor writes plain Markdown, with wiki links and headings, into a folder you choose. Open the same files in any other editor tomorrow. Nothing is locked, nothing is uploaded, and a note you save is in memory a moment later.

Already writing in Obsidian, Logseq or a folder of Markdown? Point a vault at it. It reads, it never writes into your notes.

Connect your notes →

Free, no account

The complete OS, for free. No subscription.

You download, you launch. Nobody asks your name or your email. Everything on this page runs without a license and without credits, with a local model.

How it works

Five steps, once. Then it just runs.

01 One vault per domain "+ New Vault": work, research, home. Each a file on your disk.
02 Feed each one A folder to watch, the notes editor, a page dropped on the canvas.
03 Set the dials The "RAG attention" dial per vault. MAXIMUM protection on the private one.
04 Look at the map The Neural Map shows the shape of your notes and documents.
05 Ask All vaults, or one. The answer names its sources.

The detailed walkthrough lives in the documentation: Your first vault ↗ ·The Neural Map ↗

Download, free No account, local-first

Questions

What people ask before installing.

What is an AI second brain?: A place where what you read, write and decide is kept and can be asked. Here it is a set of vaults on your own computer: notes, documents, web pages and conversations, indexed locally, answered by the model you choose, with the source named every time.
Does it work offline?: Yes. Indexing and search run on your machine, and with a local model downloaded once, the answers do too. A cloud model is a choice you make per conversation.
Does it replace Obsidian or Notion?: No. It connects to them. An Obsidian or Logseq vault is a folder of Markdown files you give to a vault to watch; a Notion export lands the same way. It reads your notes and never writes into them.
Where are my notes stored?: In a folder you choose, as plain Markdown files, with wiki links and headings. Any editor opens them. The memory index is another file on your disk. Nothing is uploaded.
Does it remember my conversations?: Yes, by default: each conversation becomes one memory in a Conversations vault, kept current as it grows. You can switch it off per conversation, and deleting a conversation asks whether to forget its memory too.
Can it keep my private notes out of answers?: Yes. Mark that vault MAXIMUM: it answers only questions that name it, and its content is not sent to a cloud model unless you confirm it, each time. You can also mute any vault with its attention dial.
Are my notes encrypted?: They can be. AES-256 encryption at rest exists and you arm it in the settings. By default your vaults are local files in the clear, readable without us.
Is it free?: The download is the complete OS, free, with no account and no email. Everything on this page runs without a license and without credits, with a local model. Cloud credits are prepaid and optional.