founder-os

Founder OS: System Overview

The complete Founder OS: 24 MCPs, 25+ skills, 6 hooks, PARA folders, persistent memory. Run a business from one terminal.

15 min read·

System Overview

Run a Business from One Terminal

The Founder OS is a five-layer system built on Claude Code. It connects your identity, memory, tools, safety rules, and workflows into one coherent environment, so Claude always knows who you are, what you are working on, and how to act on your behalf without breaking anything.

24

MCPs connected

25+

Skills installed

6

Safety hooks

1

Terminal to rule them all

Architecture

Five Layers of the Founder OS

Each layer builds on the previous one. Together they give Claude full context, full access, and full protection.

01
CLAUDE.md sets the identity that MEMORY.md extends over time
02
MEMORY.md gives context. MCPs give Claude hands to act
03
MCPs give access. Hooks make sure that access is safe
04
Hooks protect the environment. Skills use it to execute workflows
05

Second Brain

PARA Folder Layout

The PARA method (Projects · Areas · Resources · Archives) organises your second brain so Claude can navigate it without asking where things live.

Projects/

  • client-[name]/ # per-client folder
  • ├── CLAUDE.md # client context
  • ├── memory/MEMORY.md # persistent state
  • ├── briefs/ # project briefs
  • └── deliverables/ # outputs
  • internal/ # internal projects
  • └── [project-name]/

Areas/

  • sales/ # pipeline + outreach
  • content/ # posts, scripts, hooks
  • finance/ # invoices, runway
  • ops/ # SOPs, processes
  • team/ # hiring, onboarding

Resources/

  • templates/ # reusable prompts
  • hooks/ # shell safety scripts
  • skills/ # slash-command .md files
  • mcp-configs/ # MCP server configs
  • swipe-files/ # inspiration library

Archives/

  • closed-clients/ # completed engagements
  • old-projects/ # shipped work
  • deprecated-skills/ # retired automations

Setup

Wire-it-up Checklist

Eight steps to go from a bare terminal to a fully operational Founder OS.

01

Create your CLAUDE.md with business context, client list, and behaviour rules.

02

Create MEMORY.md at ~/.claude/projects/.../memory/MEMORY.md and seed it with one insight.

03

Install the 6 hooks: chmod +x each one and register them in .claude/settings.json.

04

Connect the 8 core MCPs: Attio, Slack, Notion, Unipile, Anysite, GitHub, n8n, Filesystem.

05

Set up your PARA folder structure and alias the second brain in CLAUDE.md.

06

Install 5 priority skills: /cockpit, /personalized-value-outreach, /meeting-to-html, /plan, /go.

07

Run sheal check to confirm environment is healthy before your first real session.

08

Open a session and run /cockpit to verify all layers connect correctly.

Execution Flow

How the Layers Work Together

1

Claude reads CLAUDE.md + MEMORY.md

Context loaded. It knows who you are, what you are working on, your preferences, and your corrections from past sessions.

2

Hooks protect the session

Safety active. Destructive commands are blocked, sensitive files are guarded, every edit is backed up and logged.

3

You trigger a skill

Claude uses MCPs to act on your tools: reads your CRM, drafts a message in your voice, sends it, and logs the action. One command, full execution.

Outcomes

What You Get

90 sec

to retrieve any context

No more hunting through tabs or re-pasting notes. Claude knows where everything is.

40 min

of morning orientation eliminated

Start working immediately. No re-explaining who you are or what happened last week.

Zero tabs

to check during a client call

CRM, pipeline, notes, and history: all in the terminal, all accessible in one command.

1 command

triggers any workflow

From prospect research to sent DM in under 90 seconds. No manual steps.

Five layers. One terminal. Your entire company: organised, connected, and ready for Claude to navigate.

CLAUDE.md knows you. MEMORY.md remembers. MCPs act. Hooks protect. Skills execute.

Want this running in your stack?

AY Automate builds AI automation systems for production teams.