Sovereign Brain A local AI second brain
that runs on your machine.
Sovereign Brain is a private, offline AI assistant for your ideas — a self-hosted second brain powered by a local LLM. You talk; it captures and organizes everything into plain Markdown files that live on your disk. No cloud. No subscription. No one reading your thoughts but you.
Watch a real session
A real terminal session onboarding on my laptop, played back at 2×.
100% private
Runs fully offline on local inference. Nothing is sent to any server, ever.
Plain Markdown
Every note is a .md file you own — open it in Obsidian, VS Code, or any editor.
Skill-driven capture
Just talk. Installable skills teach the agent to file each thought, todo, and decision in a consistent place — capture is frictionless.
Free & open
Public-domain template. No lock-in, no account. Fork it and make it yours.
How your local AI second brain works
You dump
Thoughts, ideas, TODOs — throw them in freely. The agent automatically organizes the information into simple Markdown files.
Brainstorm together
You and the agent work through problems together, leveraging your existing context to develop real plans.
You execute
You know what to work on — the minimum work required for maximum return — and can quickly iterate.
What's inside the template
One file to start. Add optional skills with a single command, and your data folders write themselves as you talk.
my-brain/ ├── AGENTS.md # who your brain is + how it works │ ├── skills/ # optional — add 16 workflows in one command │ ├── todos/ │ ├── daily-thoughts/ │ ├── decisions/ │ └── … # projects, people, goals, habits… │ └── (your data) # folders appear here as you talk ├── daily-thoughts/ # 2026-06-27-shower-idea.md ├── todos/ ├── projects/ └── session-notes/ # distilled brainstorms, dated
What you'll need
Sovereign Brain runs a real language model locally — and the bar just dropped: a modern 9B model now runs on a 16 GB machine.
llama.cpp plus an open model that fits your RAM — Qwen3.6-35B-A3B on 32 GB+, or the lighter Ornith-1.0-9B on 16 GB.pi agent.NEW TO LOCAL AI? SEE HOW IT WORKS ABOVE, OR THE BLOG FOR SETUP GUIDES.
Common questions
Is my data private?
Yes — it runs offline on your machine, nothing is uploaded to any server.
Does it cost anything?
No. The template is free and released into the public domain.
Can I use my own editor?
Yes — it's plain Markdown, so Obsidian, VS Code, and any editor work.
How is it different from ChatGPT or Notion AI?
Those run in the cloud on a subscription. Sovereign Brain runs locally, keeps data on your disk, works offline, and costs nothing.
From the blog
JUN 272026 Ornith 1.0 9B: The Local Brain Bar Just Dropped to 16GB
JUN 042026 Gemma 4 12B: All Hype, No Substance
JUN 012026 Best Small Local Models for Your Sovereign Brain All articles →