AI OS

Simple name, on purpose.

LLMs are powerful but unreliable. They hallucinate, forget, and treat every conversation like it's the first.
AI OS is the architecture layer that fixes that — memory, identity, learning, and control.
Everything runs locally. Your data never leaves your machine.

Live Status

Honest numbers pulled from the running system at build time, not marketing claims.

Commit 6485b00 · 161 commits total
Last change demo: ExecStart env wrapper guarantees override of EnvironmentFile (2026-04-22 23:57:50 -0400)
Thread modules 8 · 27 tools in the form registry
Background loops 13 defined (13 built-in + 0 custom) · none auto-started — user opts in per loop
Identity facts 34 stored about the user + their world
Philosophy (values) 7 written values in value_system.machine
Concept graph 4,760 concepts · 653,496 links
Self-reflections 100 meta-thoughts logged
Open goals 19 proposed or approved
Sensory bus 6 events recorded · 0/24 sources consented (off by default)

generated 2026-04-23T03:58+00:00 UTC

Try the Demo

This is the real app running in demo mode — click around, explore every page.

AI OS — Demo

What Makes It Different

  • Persistent Memory

    Remembers conversations, preferences, and context across sessions. Background loops extract and consolidate facts automatically.

  • 100% Local

    Runs on your machine via Ollama. No cloud, no tracking, no data collection. SQLite database you can inspect and export.

  • Real Architecture

    Not a chatbot wrapper. Six thread modules, 10 background loops, structured state, and a proper eval harness.

  • Self-Improving

    Generates its own fine-tuning data from interactions. Train on Apple Silicon with MLX. Your agent gets better over time.

Get Started in 5 Minutes

Download the macOS installer or clone from source. Runs on Mac, Linux, and Windows.

Requires macOS 12+ · 8GB RAM minimum · Apple Silicon recommended

AI OS

From Source

git clone https://github.com/allee-ai/AI_OS.git
cd AI_OS
pip install -r requirements.txt
python3 -m uvicorn scripts.server:app --host 127.0.0.1 --port 8000