Getting Started
cognitiveOS is an AI-powered filesystem scaffold for developers with executive dysfunction. Your AI agent reads your project state automatically, so you open your laptop and know exactly what to do, instead of spending 30 minutes remembering where you were.
Free. Open source. MIT. No servers, no accounts, no database. Just markdown files on your machine.
Install
Run this inside any project:
cd your-project
npx cognitiveos init
Answer 3 questions. Done in under 60 seconds. Open your AI agent and start working. It already knows your context.
What it generates
your-project/
├── AGENTS.md routing table, your agent reads this automatically
├── CLAUDE.md same, for Claude Code
├── STATE.md your current-state snapshot: focus, blockers, open loops
├── brain-dump/ capture everything, no filter
├── queue/ what's next, already sorted
├── focus/ the ONE thing right now (one task, enforced)
├── projects/ active work, max 3
├── ideas/ captured, not committed
├── someday/ not now, not never
└── sessions/ auto-logged history
Each folder is one cognitive mode. Your agent reads STATE.md at session start and updates it at session end, via hooks, automatically. You never run a "save" command. You never maintain anything.
This is not a productivity system. It's a cognitive prosthetic. A productivity system asks you to maintain it. This maintains itself.
Your first session
- Run
npx cognitiveos initin your project and answer the 3 questions. - Open your AI agent (Claude Code, Codex, Antigravity, Cursor). It reads
CLAUDE.md/AGENTS.mdand loads your context automatically. - Run
cognitiveos startany time to print Mission Control, exactly where you left off, in your terminal. - Capture stray thoughts with
cognitiveos dump "..."without breaking flow. - Run
cognitiveos checkto confirm everything is wired correctly.
Next steps
- Commands: full reference for
init,start,dump,check. - The Zones: what each folder means and when to use it.
- Agent Setup: wire your specific AI agent.
The bet: structure beats willpower.