AIS-OS is an AI Operating System starter kit for Claude Code that packages an /onboard, /audit, and /level-up skill set with the Three Ms framework. Released under MIT by nateherkai/AIS-OS, the project turns Claude Code into a persistent automation layer intended to observe and act while you are away from the desk. It pairs with a companion masterclass and targets the AI Automation Society community, though the kit is generic enough for anyone building automations. The repository has 67 stars and frames progress around felt outcomes rather than formal KPIs.

Core features

The kit centers on two frameworks and three recurring skills. The Three Ms (Mindset, Method, Machine) structure operator thinking, while the Four Cs (Context, Connections, Capabilities, Cadence) organize system architecture. Concrete components include:

  • /onboard interview that personalizes the AIOS to your business and context.
  • /audit skill for recurring review of constraints, processes, and autonomy levels.
  • /level-up skill that walks through the Three Ms weekly and reinforces workflows over agents.
  • references/3ms-framework.md with full definitions of Mindset, Method, and Machine.
  • Layered architecture aiming for Context (fresh Claude sessions can describe the business without browsing), Connections (live calendar and task queries without manual paste), Capabilities (phrase-triggered multi-step artifacts), and Cadence (unattended runs).

Success indicators described in the README emphasize observable shifts: teammates receiving faster, sourced answers from the AIOS; reduced context-switching as the AIOS becomes the default surface for thought work; and business facts leaving your head into retrieval so you hold questions instead of memory.

Getting it running

AIS-OS is delivered as a starter kit for Claude Code rather than a self-contained server, so installation occurs within the Claude Code environment you already operate. Clone the repository and reference its files from your Claude Code workspace:

git clone https://github.com/nateherkai/AIS-OS.git
cd AIS-OS

There is no pip install, no npm install, and no Docker container provided. The kit assumes Claude Code is available and that you will load the included prompts, frameworks, and skill definitions into the Claude Code session or workspace where you run automations. Configuration proceeds through the /onboard interview, after which /audit and /level-up can be invoked as recurring commands. Dependencies beyond Claude Code itself are not listed in the project.

Who this is for

The README identifies solopreneurs, small business operators, managers, creators, and AI consultants as the primary audience. It suits people who want Claude Code to retain context, execute workflows, and reduce daily retrieval work without building bespoke agents. The kit explicitly favors workflows over agents, Lego-style composition, and validation chains that tolerate boredom. If you expect turnkey software with a server binary or cloud dashboard, this is not that; it is a lightweight scaffold for persistent Claude Code use.

How it compares

AIS-OS differs from general agent platforms that bundle hosting, execution queues, and UIs. It does not compete with self-hosted app containers or codebases that expose REST APIs and database backends. Compared with heavier automation stacks, it is thinner, relies on Claude Code for runtime, and emphasizes personal onboarding and recurring thinking skills rather than programmable triggers. Alternatives offering visual builders or multi-model orchestration provide more surface area; AIS-OS provides a narrower, opinionated layer tied to the Three Ms and Four Cs.

The project remains a prompt-and-process kit licensed MIT, maintained at nateherkai/AIS-OS, and aligned with a design litmus test that requires observable, desk-free outputs.

View on GitHub →