Evolving Brain — Template

A template for building your own AI-native second brain / Life OS. Fork it, make your fork private, run the setup, and let it grow.

This is the public template. Do not fill in your actual identity, goals, or journal entries here. Fork it first, go private, then onboard.

One-command setup (after you fork and clone)

git clone https://github.com//.git evolving-brain
cd evolving-brain
bash scripts/setup.sh

That script verifies your tools, checks the bundled Obsidian plugins, creates Vault/.env from the template, and opens the vault in Obsidian. When Obsidian prompts “Trust author and enable plugins?”, click YES.

You get three plugins pre-installed and enabled:

• internetVin Terminal — embedded terminal in Obsidian
• Agentfiles — browse and edit AI agent skill / command files from inside Obsidian
• Excalidraw — sketch diagrams and hand-drawn mind maps

All three are MIT-licensed and bundled directly in .obsidian/plugins/ so a clone gives you a working environment with zero plugin hunting.

Then, inside Claude Code, run /brain-onboard (or /brain-discover to scan your machine for markdown sources first).

What this is

A markdown vault designed to be the source of truth for a personal AI assistant that actually remembers you. Capture tools (Heptabase, Apple Notes, voice memos, meeting notetakers, Claude on your phone) drop raw content into an .inbox/. A processor runs on a schedule, synthesizes it into living logs and entity pages, and commits back. A nightly dream cycle enriches entities, fixes broken links, and surfaces contradictions while you sleep. Any Claude surface can read the vault via a thin MCP server.

The core insights:

You are a writer compiling, not a filing clerk. The question is never “where do I put this fact?” It is: “what does this mean, and how does it connect to what you already know?” — Farzaa’s wiki-gen-skill

Every page is a dossier. Compiled truth on top (your current best understanding, rewritten when evidence changes). Append-only timeline on the bottom (the evidence trail that never gets edited). — Inspired by garrytan/gbrain

See ATTRIBUTION.md for the full credit list.

The three-folder core

• 0 - Identity — who you are (core, stable, changes rarely)
• 1 - Aspirations — who you are becoming (goals, habits, projects, open problems)
• 2 - Live Logs — the raw signal of your life (actions, decisions, emotions, learning, reflections, plus Claude’s own action log)

Everything else (3 - Daily Journal through 9 - Operations, People, Onboarding, Vault, Archive, scripts) is scaffolding that feeds into or operates on the three core folders above.

How to use this template (step by step)

1. Fork + go private

gh repo fork  --clone=false
gh repo edit / --visibility private
gh repo clone /
cd 

Do not skip the private step. The moment you fill in SOUL.md, this becomes the most personal data you own.

2. Run one-command setup

bash scripts/setup.sh

See the top of this README for what it does.

3. Run the onboarding slash command in Claude Code

/brain-onboard walks you through:

1. Discover data sources — optionally run /brain-discover first to scan your machine for markdown-heavy folders.
2. Dump data — drop exports from your existing tools into .inbox/onboarding//.
3. Generate the wiki — /brain-quick-wiki builds a Wikipedia-style static site.
4. Answer identity questions — fill in Onboarding/01 and 02 when you’re ready.
5. Wire a connector — pick one source (recommend: meeting notetaker).
6. Schedule the processor — GitHub Actions on a cron is the simplest path.

4. Daily use

• Capture on the fly: /brain-capture "the thing I want to remember"
• Morning briefing: /brain-brief
• Check state: /brain-status
• Weekly review: /brain-review weekly
• Regenerate wiki: /brain-quick-wiki
• Clean up stray files: /brain-tidy
• Check upstream sources: /brain-sync

Runtime visibility (9 - Operations/)

One of the things this template gets right that other “AI second brain” setups miss: a single place to see what’s running.

• 9 - Operations/workflows/README.md — registry of every workflow, status, schedule, scope, last run.
• 9 - Operations/connectors/README.md — every MCP / poller / watch folder feeding .inbox/.
• 9 - Operations/skills/README.md — bundled Obsidian plugins + Claude skills the vault depends on.
• 9 - Operations/schedule.md — calendar-style view.
• 9 - Operations/runs/ — append-only run log per month.
• 9 - Operations/upstream/ — pinned copies of files from prior-art projects (gbrain, etc.) with a /brain-sync mechanism for pulling in upstream improvements.

Credentials (Vault/)

• Vault/.env — the big bolt. Gitignored.
• Vault/.env.example — committed template.
• Vault/key-inventory.md — human-readable catalog. No values.

Workflows never hardcode secrets.

Entity pages (the dossier pattern)

Every People/, 5 - Projects/, 6 - Areas/ page follows the compiled-truth + append-only timeline structure borrowed from gbrain’s recommended schema. Top is current best understanding (rewritable). Bottom is append-only evidence trail (never rewritten). When new evidence contradicts the top, the processor flags it — it never silently resolves. Human always decides.

Folder map

Status

This template ships with Phases 0 and 1 complete. Phases 2-9 are your job after forking. See MASTER PLAN.md for the full build plan.

License

MIT. See LICENSE.