Skip to content

LifeOS

Overview

LifeOS (Daniel Miessler) is an AI-powered personal operating system designed to help a user move from a "Current State" to an "Ideal State" in both life and work. Rather than being a standalone agent, it functions as a "general-purpose hill-climbing harness" — a configurable layer that runs on top of an existing AI coding platform (primarily Claude Code, though designed to be harness-agnostic) and adds persistent memory, intent-driven task routing, and self-modifying capabilities so the underlying model's raw reasoning power becomes personalized to the user's specific goals, preferences, and decision history. It sits in the same "personal AI agent" category as Hermes Agent, OpenClaw, and OpenHuman, distinguished by being distributed as a single self-contained Claude Code skill rather than a standalone application.

Key Concepts / Architecture

LifeOS bundles several subsystems into one unified skill:

  • Cortex — memory architecture: knowledge graph, work history, and accumulated learnings.
  • Synapse — input routing and task classification; dispatches incoming requests to the appropriate workflow.
  • Atlas — live asset management and relationship tracking.
  • Ledger — change tracking and audit logs.
  • The Algorithm — a seven-phase operating loop: OBSERVE → THINK → PLAN → BUILD → EXECUTE → VERIFY → LEARN.
  • ISA System (Ideal State Artifact) — a structured document capturing the user's goals across multiple identity lenses.
  • Pulse — a unified daemon providing voice interfaces, scheduling, dashboards, and optional messaging bridges.
  • Skills Library — 49+ specialized workflows spanning research, writing, security, and other domains.

Architecturally, LifeOS positions itself as the operational layer above raw AI capability: the underlying harness (e.g., Claude Code) supplies reasoning and tool execution, while LifeOS supplies continuity — context, preferences, decision history, and goal-state alignment — across sessions.

Technology Stack

  • Language: TypeScript and Bash
  • Runtime: Bun
  • Primary platform: Claude Code (designed to be harness-agnostic)
  • Distribution: single self-contained skill, installable via a prompt or curl | bash
  • Persistence: Git-backed version control and local file storage

Notable Design Patterns

  • Constitution-as-code — identity and behavioral principles load as system prompts.
  • Capability awareness — the installer detects which optional features a user wants and provides specific remediation on failure.
  • Recovery-by-design — every component can be rolled back; user customizations in a dedicated USER directory survive upgrades.
  • Community-integrated development — public PRs are ported into a private source tree, preserving contributor credit while maintaining release stability.

Licensing

MIT License — free and open source, with explicit commercial-use freedom.

Suitable for (Pros)

  • Individual power users of Claude Code (or another supported harness) who want persistent, cross-session memory and goal tracking without adopting a separate standalone agent platform
  • Users who want a self-contained, auditable, git-backed personal knowledge/memory layer rather than a hosted SaaS memory service

Limitations (Cons)

  • Tied in practice to Claude Code as the primary supported harness, despite the harness-agnostic design goal
  • Single-user, local-first design — not intended for team or multi-tenant deployment
  • As a skill/harness layer rather than a hosted platform, it depends on the capabilities and availability of the underlying AI coding platform it runs on

See Also

References