Your AI Homestead

Discover, track, and serve any local LLM from any engine — with one CLI.

$ homestead discover

Scanning Ollama... Done (7 models)

Scanning HF Hub... Done (8 models)

Scanning GGUF files... Done (19 models)

Scanning MLX... Done (1 model)

Registry updated. 41 models tracked.

Zero-Config Discovery

Find every model on your machine — Ollama, HuggingFace cache, GGUF files, MLX, and running engines — in one SQLite registry.

Cross-Engine Serving

Serve any model with a single command. Homestead auto-selects the best engine — Ollama, llama.cpp, MLX, or Apple Core AI.

Audit Trail

Every model load, unload, and error logged to SQLite and streamed via SSE. Prove no data left your machine.

Simple by Design

Step 1

Install Homestead

Requires Bun. Clone the repo and install dependencies.

# Install
npm install -g homestead
git clone https://github.com/olanigan/homestead.git
cd homestead
bun install

Step 2

Discover Models

Scan your machine for LLMs. Homestead finds models across Ollama, HuggingFace cache, GGUF files, MLX, and running engines.

# Full machine scan
homestead discover

Step 3

Serve & Chat

Launch an OpenAI-compatible API server for any model in your registry with auto-selected engine.

# Start serving
homestead serve qwen