What AI harness does LifeOS require?

Any capable agentic AI coding harness works in principle, since LifeOS is built on universal primitives like hooks, skills, and context files. Claude Code is the most tested path because that's what the maintainer builds and runs on.

Is LifeOS self-hosted?

Yes. It installs locally into your existing AI harness setup (for example, your `~/.claude` directory) and runs through that harness rather than a hosted service.

What license does LifeOS use?

MIT license, so it's free to use, modify, and redistribute.

How does LifeOS differ from Fabric?

Fabric is a collection of AI prompts (patterns) focused on what to ask AI. LifeOS is infrastructure for how your assistant operates, covering memory, skills, routing, and context, and the two are designed to be complementary.

What happens if an upgrade breaks my setup?

The README recommends backing up `~/.claude` before upgrading. Your `USER/` directory is never touched by the installer, settings merge rather than overwrite, and history is git-backed so you can roll back.

Does LifeOS support local models?

Not yet. Local model support (Ollama, llama.cpp) is listed on the roadmap along with granular model routing, remote access, and outbound calling, but isn't shipped as of this README.

LifeOS: An AI harness layered on your coding

LifeOS is an AI harness that sits on top of an existing coding agent (Claude Code, Cursor, Codex, or similar) and turns it into something that knows who you are, what you care about, and where you want to go. It's built for people who already use an AI coding harness and want that harness to carry persistent context across sessions instead of starting from zero every time. The project frames its purpose around moving from a "Current State" to an "Ideal State" in both life and work, using memory, routing, and a skill system to make that transition concrete.

Unlike a prompt library or a single agent configuration, LifeOS installs as a self-contained skill bundle that includes research, security, writing, art, and other capability sets. It's harness-agnostic by design, built on primitives like hooks, skills, context files, and agentic routing rather than tying itself to one vendor's API. The maintainer builds and tests primarily against Claude Code, so that's the most reliable path today, but the core concepts are meant to port to any capable agent.

Key features

Persistent memory: your digital assistant (DA) retains past sessions, decisions, and learnings instead of forgetting context between conversations.
Skill system: a bundled library covering research, security, writing, art, and more, browsable and installable as one package.
Context files (TELOS): goals, contacts, and preferences are stored so you don't re-explain yourself every session.
Intelligent routing: natural-language triggers (like "research this") kick off the matching workflow automatically.
Self-improvement: the system can modify itself based on what it learns from use.
Safe customization: a dedicated USER/ directory holds personal config that installers and upgrades never overwrite.
Git-backed recovery: version control and preserved history make it possible to roll back changes or restore DA memory after a mistake.

Ideal use cases

LifeOS fits developers and power users who already run an AI coding harness daily and want it to carry long-term context, personal goals, and recurring workflows (research, writing, security tasks) without re-explaining themselves each session. It suits people building a personal operating layer around AI for both professional projects (app building, business tasks) and personal life management.

It's not a good fit if you're looking for a standalone AI model, a simple prompt collection, or a tool that works without a capable AI coding harness already installed. If you want something lightweight with no setup overhead, or you're not using (and don't want to use) an agent like Claude Code, this adds friction rather than value. It's also not the right choice if you need local-model-only operation today; that support is on the roadmap but not yet shipped, along with remote access and external notifications.

Installation

LifeOS installs itself through your AI agent rather than a traditional package manager step. You need a capable AI coding harness (Claude Code is the most tested) and bun installed first.

The simplest path is to paste a prompt into your agent:

Read https://ourlifeos.ai/install and install LifeOS for me.

Your AI reads the install page and walks through setup, asking permission before making changes.

If you prefer the terminal and use Claude Code on macOS or Linux, there's a one-line installer:

curl -fsSL https://ourlifeos.ai/install.sh | bash

Before upgrading an existing install, back up your Claude config directory:

cp -r ~/.claude ~/.claude-backup-$(date +%Y%m%d)

Running the installer again on an existing setup detects the prior install and merges updates intelligently, leaving your USER/ customizations and hooks untouched.