Guide · Open source
How to use LifeOS: install, set up and get started
Updated · written with AI from the project's own documentation and checked against our catalogue data
LifeOS is an open-source (MIT) set of files, skills, and routing rules that sits on top of an AI coding harness such as Claude Code. It stores your goals, people, and preferences so the AI can keep working toward them without you re-explaining context each time. This guide is for people who already use an AI coding harness and are comfortable with a terminal; by the end you will know what you need, how the install works, and how to keep your setup safe.
Quick start
- 1Check prerequisites. Confirm you have a capable AI coding harness such as Claude Code and bun installed.
- 2Choose an install path. Use the AI prompt for any supported harness, or the one-line installer if you use Claude Code on macOS or Linux.
- 3Run the install. Paste the prompt into your harness or run the curl installer in your terminal, and approve each change the AI asks about.
- 4Back up before changes. Copy your Claude configuration folder to a dated backup before any upgrade.
- 5Put your goals in USER/. Store your customizations in USER/, which the README says the installer does not overwrite.
- 6Try a skill. Ask your harness a plain request such as "research this" and check that the matching workflow runs.
What you need before installing LifeOS
LifeOS needs a capable AI coding harness. The README names Claude Code as the most-tested path and lists Cursor, Codex, and Hermes as examples of agents you can paste the install prompt into. You also need bun, the JavaScript runtime the project lists as a prerequisite.
The project is written in TypeScript and Bun is part of its setup, so the install is aimed at developers. If you do not already use an AI coding harness, the README says LifeOS is not for you.
- A capable AI coding harness, such as Claude Code
- bun
- Comfort with command-line setup
Install LifeOS with an AI prompt
The README says LifeOS is installed by an AI. Paste the install prompt into your AI coding harness and it reads the install page and walks you through setup. According to the README, the AI asks permission before it touches anything.
This is the path that works with any supported harness, not only Claude Code.
Read https://ourlifeos.ai/install and install LifeOS for me.Install LifeOS from the terminal on macOS or Linux
If you prefer the terminal, the README gives a one-line installer for Claude Code on macOS and Linux. Run it in a terminal that has access to your Claude Code setup. The README does not list a Windows equivalent of this command.
After the installer finishes, the README says re-running it later detects existing installations and merges changes rather than overwriting them.
- The one-line installer is for Claude Code on macOS/Linux only, per the README
- Re-running the installer is the documented way to update an existing setup
curl -fsSL https://ourlifeos.ai/install.sh | bashBack up your setup before upgrading
Before any upgrade, the README recommends copying your Claude configuration folder. The command below creates a dated backup in your home directory.
The README says your customizations in USER/ are never touched by the installer or upgrades, and the installer only updates identity and version fields in settings. Hooks, statusline, and custom config are preserved. It also suggests keeping your setup under version control so you can roll back.
- Keep your customizations in USER/, which the README says the installer does not touch
- Version-control your setup so you can roll back
- Your DA's memory is preserved across mistakes, according to the README
cp -r ~/.claude ~/.claude-backup-$(date +%Y%m%d)Understand the core ideas: Current State, Ideal State and skills
The README describes one central concept: moving from your Current State to your Ideal State. LifeOS stores your goals, contacts, and preferences so the harness has that context for every task.
The README lists several features that LifeOS adds on top of a harness: persistent memory, custom skills, your context, intelligent routing (for example, saying "research this" triggers the matching workflow), and self-improvement. It says there are 97 public skills, which you can browse on the project site.
- Persistent memory of past sessions, decisions, and learnings
- Custom skills for the tasks you repeat
- Routing from plain requests to the right workflow
- Browse the public skills at ourlifeos.ai/skills
Know the current limits
LifeOS is not a polished consumer app. Install is driven by an AI and a terminal. The README lists local model support (Ollama, llama.cpp) only on its roadmap, so today it depends on a hosted harness such as Claude Code.
Remote access, outbound phone calling, and external notifications for email, Discord, iMessage, and Slack are also listed as roadmap items, not current features.
- Local models are roadmap items, not current support
- Remote access is a roadmap item
- Outbound calling and external notifications are roadmap items
What people use it for
- Keeping goals, contacts, and preferences available to an AI coding harness across sessions
- Running reusable research, writing, or publishing skills from one setup
- Building a personal AI assistant that refines its own workflows over time
- Working on app building, business planning, or creative projects with stored context
Frequently asked questions
Is LifeOS free?
LifeOS is licensed under MIT, and the README describes it as open source. The README does not give pricing for the AI harness it runs on, so check the terms of whichever harness you use, such as Claude Code.
What do I need to run LifeOS?
The README lists a capable AI coding harness such as Claude Code and bun as prerequisites. It also says Claude Code is the most-tested path. Users need to be comfortable with TypeScript, Bun, and command-line setup.
Can I use LifeOS without Claude Code?
Yes, according to the README, LifeOS is harness-agnostic and the install prompt works with Cursor, Codex, Hermes, or any capable agent. Claude Code is the path the author builds and tests on.
Does LifeOS work with local models like Ollama?
Not yet. The README lists local model support, including Ollama and llama.cpp, only on its roadmap. Privacy-focused users who need fully local models today should choose another tool for now.
How is LifeOS different from Fabric?
The README says Fabric is a collection of AI prompts for specific tasks, while LifeOS is infrastructure for memory, skills, routing, and context. The README says the two are complementary and many LifeOS users integrate Fabric patterns into their skills.
Will an upgrade erase my customizations?
The README says your files in USER/ are never touched by the installer or upgrades, and the installer only updates identity and version fields in settings. It still recommends backing up your configuration folder before any upgrade.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
LifeOS suits developers who already use an AI coding harness such as Claude Code and want it to remember their goals, contacts, and preferences across sessions. It is a poor fit if you want a polished app with no setup, or if you need fully local models today, since local model support is only on the roadmap.
Install options & downloads