HangarMachine-aware local AIRepository
Local AI, fitted to your machine

Set up your local AI stack with one skill.

Choose Codex or Claude Code. Your agent measures your Mac, Windows PC, or Linux box, finds what is installed, chooses a local LLM that fits, and launches your private Hangar console at localhost:8003.

Pick your coding agent, then paste the prompt
Use the Hangar setup skill at:
https://github.com/alibad/hangar/blob/main/.agents/skills/hangar-setup/SKILL.md

Set up Hangar on this machine. Measure the machine first, choose a local model that fits, configure only services you verify, and prove the console works locally before you finish.
Open Codex skill

Open source · local by default · no account required. This website is the launch point, not the console running on your computer.

Measures RAM + GPU
Uses what is installed
Chooses a fitting local LLM
Verifies before it finishes

The local console

See what Hangar gives you.

These are real captures from configured machines. Hangar changes with the hardware and services it discovers—so a Mac and a GPU workstation will not pretend to have the same capabilities.

Hangar home showing machine capacity, ready services, workstreams, and a local model fit check

One view of your machine

See memory headroom, what is ready, recent requests, and whether a workflow can run before you launch it.

Hangar model catalog filtered to models that fit an Apple M5 Max machine

Models sized to your hardware

Compare installed and available models against the RAM, GPU, and storage this computer actually has.

Hangar services control center showing Ollama, the service manager, and agent access configuration

Control every local service

Start, stop, inspect, and connect services to Codex or Claude Code without hunting through terminals.

Hangar Image Studio on a configured NVIDIA host showing a locally generated lighthouse image

Create with the stack

Shown on a configured NVIDIA host named BeTenshi

Use the same console for chat, local image generation, speech, vision, and 3D when the machine supports them.

1. Inspects your machine

Detects macOS, Windows, or Linux, available memory, accelerators, Ollama, and reachable local services.

2. Configures Hangar

Creates a profile for this computer and wires only the models and capabilities it can prove are available.

3. Proves it works

Runs the tests, starts the console, checks the host profile, and verifies the local model service responds.

Prefer to install it yourself?Show the manual steps

Install the basics

Install Git, Node.js 22+, and Ollama, then clone Hangar.

Clone and install
git clone https://github.com/alibad/hangar.git
cd hangar
npm install

Add a local model and start

Qwen3 8B is a sensible first local text model. The setup skill can choose a better fit after checking your hardware.

Model and console
ollama pull qwen3:8b
node scripts/detect-host.mjs
npm run dev

Your models stay on your machine

This public site is only the setup guide. The working console and its machine APIs run locally; the public Human Quest domains cannot reach them.

View source