second-brain

Save UI inspiration from any app. A vision model curates it into a private wiki and turns your taste into a DESIGN.md your coding agent builds from.

• Save from anywhere. Share a post from X, Instagram or LinkedIn, any web page, a screenshot or a photo, from your phone or your computer.
• Curated, not dumped. A vision model reads the frames, measures the palette from pixels and files each capture under a topic: design, features, tools or practices. Code counts your recurring choices into a taste profile.
• Your DESIGN.md. The taste, or any saved style, becomes a DESIGN.md: tokens derived by code, contrast-checked, lint-clean. getdesign.md gives you Stripe's DESIGN.md; second-brain gives you yours.

This repository is the engine. Your wiki lives in a separate private repository. What sets it apart and where it's going: PRODUCT.md · ROADMAP.md.

What you get

All of it lives in your private wiki repository, written by the worker and read by your agent.

Output	What it is
The wiki	One page per topic, one source note per capture (with up to four frames of what it looked like) and an index, kept tidy by a weekly lint pass
taste.md	Your recurring choices across captures, as a default look and motion
DESIGN.md	The taste as design tokens, with a dark (or light) twin when your captures split; any style topic gets its own
AGENTS.md	The ideas of your practices topics as rules a coding agent follows
stack.md	The tools you saved, by job: used, to try or stale
usage.md	What was built from which captures, so what proves useful stands out

Quick start

You need a GitHub account with gh logged in and an API key for the model: a free Ollama key by default, or OpenAI, Anthropic, Gemini or OpenRouter. Capture from an iPhone or Mac with Shortcuts, an Android phone or any computer with a browser.

1. Get the engine:

   git clone https://github.com/albegosu/second-brain ~/Developer/second-brain
   cd ~/Developer/second-brain
   

2. Set everything up. It asks for the model provider and its key, creates your private wiki repository and a private inbox repository next to it, and wires them together:

   bin/setup
   

   It stops once to let you create a capture token on GitHub (a page it opens for you: select only the inbox repository, Contents: Read and write).

3. Connect your devices with what it prints: on an iPhone or Mac, open shortcut/Save to second-brain (GitHub inbox).shortcut and answer its two questions; on Android or a computer, scan the QR code (or open the link) and install the capture page.

4. Optionally, subscribe to the printed topic in the ntfy app to get a push when each capture is filed.

5. Share a post or a page. A few minutes later it's in your wiki, and Claude reads it through the second-brain skill.

That's all. The docs explain the pieces, how to set them up by hand, and how to run the worker on your own computer instead.

Capture from