LifeOS is a personal AI harness layered on top of an AI coding agent. It records who you are, what you care about, and where you want to go, then uses that context so the agent can help with any task, from building apps and starting a business to creative projects and personal goals.
The system is organized around one loop: describe your current state and your ideal state, then close the gap in checkable steps. Each task follows four moves (articulate what done means as criteria, climb toward it, verify the step with evidence from a tool, and learn by folding the result back into the system).
LifeOS is installed by an AI: you hand the install page to a capable coding agent and it walks the setup, asking permission before it changes anything. It is harness-agnostic by design, built on hooks, skills, context files, and agentic routing in TypeScript and Bash, with Claude Code as the most-tested host. It installs as one self-contained skill that bundles a library covering research, security, writing, art, and more.
LifeOS sits alongside Fabric, a collection of task prompts: Fabric covers what to ask an AI, while LifeOS covers how a personal digital assistant operates (memory, skills, routing, context, and self-improvement). Local model support, per-task model routing, remote access, outbound phone calls, and external notifications are on the roadmap.
Features
- Current to ideal state: goals are written as a current state and an ideal state, and the system works to close the gap in steps you can check
- TELOS: an interview captures your mission, goals, beliefs, and challenges, and every later task reasons against them
- The Algorithm: a four-step loop (articulate, climb, verify, learn) applied to code, writing, health, or life goals
- Evidence-based verification: each criterion closes only on proof from a tool
- Persistent memory: the assistant remembers past sessions, decisions, and learnings
- Skill library: one bundled skill with workflows for research, security, writing, art, and more
- Intelligent routing: a request such as research this triggers the matching workflow automatically
- Self-improvement: the system modifies itself based on what it learns
- AI-driven install: a coding agent reads the install page and performs the setup, preserving user customizations across upgrades
- Hook system and voice: hooks, a voice component, and a Hermes sidecar extend the host harness
