What is rag-os?

This repository is not a traditional software package you install. It is a single Markdown file called a build guide that describes how to construct a personal AI operating system. The idea is that you paste this guide into a coding agent of your choice, such as Claude Code or Cursor, and the agent builds the system with you. Along the way it stops and asks you questions about which chat app to use, which AI model vendor to pick, which database to store data in, and how much autonomy the agent should have. Your answers get recorded in a decisions file, and the system is built in stages with each step tested before it counts as finished. The system itself is an always-on agent that can be reached from your phone, continues working after you disconnect, and remembers what it learned across sessions. Its memory lives in plain Markdown files stored in git, so you can read and audit everything the agent knows rather than trusting an opaque cloud memory blob. A small supervisor program coordinates the pieces, and nothing the AI model remembers during a session is relied on for permanent storage. If the system crashes or restarts, no lasting state is lost. The architecture has twelve parts, including adapters for chat apps and command line tools, a security model that checks every file action against path boundaries, and a budget controller that pauses autonomous work before it drains your spending quota. The guide also covers a knowledge wiki the agent maintains as a librarian, scheduling for backups and health checks, and an operations console for auditing what the agent did. The guide includes an honest comparison of similar projects like Letta, mem0, Hermes Agent, and AutoGPT. It acknowledges that most of the individual ideas already exist as mature products. The value here is having all the pieces assembled into one transparent system that you own and run on your own hardware. A starter folder provides a runnable reference for the first two milestones. The full README is longer than what was shown.