One long-term memory. Many AI agents. Local-first. Tool-agnostic.
SuperBrain is a ready-to-use template for an AI-native second brain: a shared Obsidian-compatible vault plus a suite of skills that any agent (Claude Code, Claude Desktop, Craft Agents, Codex, Cursor, n8n, …) can read, write, and reason over.
It's built for people who use multiple AI tools and are tired of each one starting from zero.
"Your Second Brain was storage. Your SuperBrain is a colleague."
superbrain/
├── vault/ Empty-but-structured SuperBrain skeleton
├── skills/ 20 skills (core operations + Memory-Kit overlay)
├── examples/ Dummy demo notes showing the conventions
├── docs/ Architecture, workshop guide, PARA migration guide
├── LICENSE CC BY-SA 4.0 — content, skills, vault, docs
└── LICENSE-CODE GPL-3.0 — any executable code
# 1. Clone
git clone https://github.com/CoachSteff/superbrain.git ~/superbrain-starter
# 2. Create your SuperBrain vault (anywhere you like — here: ~/Obsidian/SuperBrain)
mkdir -p ~/Obsidian/SuperBrain
cp -R ~/superbrain-starter/vault/. ~/Obsidian/SuperBrain/
# 3. Install the skills globally
mkdir -p ~/.agents/skills
cp -R ~/superbrain-starter/skills/* ~/.agents/skills/
# 4. Point the skills at your vault
mkdir -p ~/.obsidian-wiki
echo 'OBSIDIAN_VAULT_PATH="'$HOME'/Obsidian/SuperBrain"' > ~/.obsidian-wiki/config
# 5. Open your vault in Obsidian. Read vault/README.md (the agent access contract).Now any agent that honours the .agents/skills/ convention can:
- Load your current context:
superbrain-orient - Ingest documents into your vault:
superbrain-ingest - Answer questions against it:
superbrain-query - Write the end-of-day summary:
superbrain-close-day - Prepare a handoff for the next session:
superbrain-handoff
SuperBrain is a three-layer system — raw sources, distilled wiki, and the schema that governs how they connect. Add a temporal overlay (daily/, context/, memory/) borrowed from the Claude Memory Kit, and you get both semantic depth and session continuity.
Full explanation in docs/architecture.md.
| Category | Skills |
|---|---|
| Core operations | superbrain-setup, superbrain-ingest, superbrain-update, superbrain-query, superbrain-lint, superbrain-status, superbrain-rebuild, superbrain-export |
| Graph & taxonomy | superbrain-cross-linker, superbrain-taxonomy |
| Ingestion variants | superbrain-data-ingest, superbrain-claude-history, superbrain-codex-history, superbrain-history |
| Session overlay | superbrain-orient, superbrain-close-day, superbrain-handoff, superbrain-compile |
| Meta | superbrain-theory, superbrain-skill-creator |
Each skill is a single SKILL.md with YAML frontmatter and instructions. Read them directly to understand what they do — they are literally the agent's prompt when triggered.
- Individuals building a cross-tool personal knowledge base.
- Teams prototyping shared agent memory (treat SuperBrain as a seed — real teams need governance).
- Trainers / workshop facilitators teaching AI-era PKM. See
docs/workshop-guide.md.
Split model:
- Content — vault skeleton, skill instructions (
.md), docs, examples — under CC BY-SA 4.0. Share and adapt with attribution; derivatives stay copyleft. - Code — any executable scripts or utilities we add later — under GPL-3.0.
The split reflects the reality that prose-heavy works (skill instructions, docs) fit CC-BY-SA better, while code fits a software licence. See both files for the full legal text.
- The three-layer wiki pattern comes from Andrej Karpathy's LLM-Wiki.
- The Memory-Kit overlay (daily / context / memory) adapts the Claude Memory Kit pattern.
- The original Second Brain methodology is Tiago Forte's.
- This template is maintained by Steff Vanhaverbeke (CoachSteff) as part of the Superworker Toolbox.
Pull requests welcome. If you add a skill, follow the shape in skills/superbrain-orient/SKILL.md — a single SKILL.md with name: and description: frontmatter is enough. Prefer clarity over cleverness; these files are read by AIs and humans both.
SuperBrain is part of the Superworker Toolbox — open-source tools for thriving in an AI-augmented workplace.