Operator OS

A framework and starter kit for running your work and projects with Claude Code as your operating team.

Seven layers, from house rules to a night shift, and one installer: /onboard sets it up on your Mac while you click answers.

Read the frameworkInstall in 15 minutes

The seven layers

Most people use AI like a search box: one question, one answer, start over tomorrow. Operator OS sets Claude up the way you'd set up a small team: house rules, shared memory, specialists, access to your tools, a night shift, and clear limits on what it may do without you.

ConventionsGrill first · popups · parallel · safetyStart here if Claude works hard but on the wrong thing.
RulesCLAUDE.md in 4 levelsStart here if Claude keeps forgetting how you like things done.
MemoryOne fact per noteStart here if you keep re-explaining context.
SkillsReusable playbooksStart here if you repeat the same kind of request.
IntegrationsConnectors, CLI, browserStart here if you copy-paste between apps and Claude.
AutomationJobs that run while you sleepStart here if you want work done before you wake up.
AccountsPersonal vs employer, NDA wallsStart here if you have an employer or client under NDA.

Underneath all seven: five ideas. Developers go further with harness settings, permissions and hooks. Operator OS by The On It Studio — theonitstudio.com Single-user licence: personal use or one business. No resale or redistribution. See LICENSE.

Five ideas behind it

The files are one implementation. Understand these and you can rebuild the system in any tool. Read the philosophy

  1. Treat Claude like a team, not a chatbot

    A chatbot answers one question and forgets you. A team knows who you are, remembers what went wrong last month, follows house rules, has specialists, and keeps working while you sleep.

  2. Every incident becomes a rule

    The system grows from mistakes, not from planning. When something goes wrong (Claude sends an email you wanted to review, a backup job dies silently, five agents collide in one folder), you don't just fix it. You write it down as a rule with the reason attached:

    Rule: External email is draft-only. Why: an email went out with a wrong figure before the operator saw it. How to apply: create a draft, show it, the human clicks Send.

  3. Decide before doing

    Most wasted AI work comes from starting before the goal is clear. So Claude grills you first: it maps the open decisions, asks them as multiple-choice popups in a few rounds, and only starts once you both agree. Facts it can look up itself, it looks up. Decisions are yours.

  4. Load context lazily

    Claude's context window is a budget. Loading everything about every project into every conversation makes it slower, more expensive and more confused.

  5. Humans hold the irreversible buttons

    Claude can draft, research, build, deploy to staging, and prepare everything. Some actions stay with you:

    • Sending anything external (email, chat to clients, posts)
    • Spending money (ads, purchases, budgets): Claude creates things paused
    • Submitting (applications, forms, payments)
    • Production systems you don't own

Two ways in

Works for people who never write code and for developers. Pick your line.

Core patheveryone, ~1 week

  1. Read 00 Philosophy (5 min).
  2. Run /onboard with Rules + Memory + Core skills + Connectors.
  3. Work normally for a week. Each time you correct Claude, say "remember this as a rule".
  4. Day 7: run /memory-prune and /weekly-planning.
  5. Add the morning briefing job once you trust the briefing you get by hand.

Advanced pathdevelopers

Everything above, plus: dev CLIs, worktrees per session, third-party skill packs, the full automation set, multi-account with VS Code profiles, hooks. Read the Advanced section at the bottom of each doc.

Meet Maya

All examples follow one fictional operator so you can see how the pieces connect:

  • Maya Chen, product manager at Acme Corp by day (NDA, separate Claude work account).
  • Side projects: Northlight Studio (design/dev agency, main income), Ridge Coffee (subscription e-commerce), Clinicly (early SaaS).
  • Sam, a part-time assistant who gets a brief every Monday.

Her filled-in setup lives in example/. Compare it with the blank templates/.

The weekly loop

The system is never finished. Fifteen minutes a week keeps it sharp:

  1. Notice friction. "I explained this again." "That job broke again."
  2. Encode it as a rule (CLAUDE.md), a memory, a skill or a job.
  3. Prune. Run /memory-prune and delete rules that no longer earn their place.

Then again next week.

Get the kit

Reading is open to everyone. The kit is $149 at theonitstudio.com/operator-os; existing invitees can still use their invite code below. Unzip it, run cd operator-os && claude, and type /onboard. Full install steps

Get the kit — $149

Zip, about 100 KB. Reading the docs needs no code.

What's inside
operator-os/
├── README.md                  ← you are here
├── .claude/skills/onboard/    ← /onboard: Claude installs the kit for you
├── docs/                      ← the framework, one doc per layer
├── templates/
│   ├── claude-md/             ← user / machine / workspace / project CLAUDE.md
│   ├── memory/                ← MEMORY.md index + note formats
│   └── settings/              ← settings.json + permissions warning
├── skills/                    ← write-professional, verify-before-done, 6 optional life-OS skills, PACKS.md (community packs, link only)
├── integrations/              ← gws-as multi-account wrapper, TOOLS.md template
├── automation/                ← job scripts, prompts, launchd templates, installer
├── multi-account/             ← account setup, skill sync, VS Code per account, git-crypt
└── example/                   ← Maya's filled-in setup