Stop paying the LLM to re-discover what matters. The Brain pre-distills the right 1.4% of your vault every session — exactly what's relevant now. They say humans use 10% of their brain. Your portable brain uses 1.4%. That's why it works.
Every AI you use is amnesiac by default. The Brain fixes three things: what your AI remembers, what it notices, and what it picks up on.
Retrieval across memory, observations, daily notes, and CRM. Your AI references prior decisions when you bring them up — without you pointing to the file. It pulls the receipts.
Silent pattern tracking. Topics that come up three times. People you keep mentioning. Quiet stress signals. Surfaced when relevant — never as a lecture, never before the moment is ready.
Session handoff written at the end of every conversation, read at the start of the next. Across days, across surfaces, across months. No more cold starts.
This is the context layer. It doesn't replace your AI — it makes any AI immediately useful. Drop it in, and your assistant knows who you are.
Monday kickoffs, daily focus, Friday reviews, quarterly rocks. Your agent detects the day type and adapts its greeting — no manual switching.
Not a sales CRM. A people manager. Tracks the whole person — birthdays, gift ideas, open loops — alongside professional context. Auto-updated from your notes.
Bash scripts run overnight — vault health scores, contact scanning, pattern detection. Free to run, no AI needed. The smart layer kicks in when you're present.
Your agent writes a handoff note at the end of every session. Next morning, it picks up where you left off — no more cold starts.
Name it, shape its tone, set boundaries. Your agent reads SOUL.md on startup and behaves the way you want — direct, warm, no-fluff, whatever.
No databases, no lock-in, no proprietary formats. Everything is .md files in folders. Works with Obsidian, VS Code, or any editor. Works forever.
A million tokens of vault, accumulated over months. The Brain delivers exactly what your AI needs this session — never the whole thing. The other 98.6% stays available, but doesn't burden the conversation.
Different days need different conversations. The agent adapts automatically.
| Day | What your agent does |
|---|---|
| Monday | Last week's results, suggested Big 3 with reasoning, project pulse, calendar, follow-ups |
| Tue — Thu | Big 3 progress, today's schedule, anything needing attention |
| Friday | Week review snapshot, then a planning conversation for next week |
| Quarter end | Big Rocks status, area check, then next quarter planning |
| Ad-hoc | Say "let's do a weekly review" any day — it loads the right context |
PARA-based folder system with agent runtime, planning cascade, and relationship intelligence built in.
Everything lands in the inbox. The agent routes it later.
If it doesn't end, it's an area — not a project.
CLAUDE.md points to the agent runtime. Any LLM picks it up.
Bash detects (free, runs overnight). LLM processes (smart, runs when you're present).
Bash scripts run on schedule — no AI tokens, no cost. They detect patterns and prepare briefings for your next session.
No accounts, no cloud, no subscriptions. Download, answer a few questions, done.
Download the zip, double-click Install Brain.command (macOS), or run bash start.sh in Terminal. Answer a few questions about yourself.
Open the vault folder in Obsidian (or any markdown editor). Your files, your folders, your data — all local, all yours.
Open the vault in Claude, ChatGPT, or any LLM. It reads CLAUDE.md automatically. Your agent introduces itself and starts helping.
Open source. Plain markdown. Works with any AI.