# ModelHub

> A macOS menu-bar app that lets you discover, download, and manage local LLMs from Hugging Face, with support for Ollama, MLX, LM Studio, llama.cpp, and vLLM.

ModelHub is a lightweight macOS menu-bar application built by Conscious Engines (Sabesh Bharathi) that centralizes local LLM management in a single ~4 MB app. It is available as a free, open-source download for Apple Silicon Macs running macOS 26 or later, currently at version 1.3.0.

## What It Is

ModelHub sits in the macOS menu bar and acts as a unified control layer for local large language models. It scans the standard Hugging Face cache (`~/.cache/huggingface/hub/`) and LM Studio's model directory (`~/.lmstudio/models/`) every time you open the menu, giving you a single list of every model on your machine. From the same interface you can browse Hugging Face, search for text-generation models, download them with one click, and remove ones you no longer need — without switching between browser tabs or terminal windows.

## How It Fits Into the Local LLM Stack

ModelHub is designed to complement, not replace, the inference tools developers already use. It writes models in the official Hugging Face cache layout — blobs, snapshots, and refs — so a model downloaded through ModelHub is byte-identical to one fetched with `huggingface-cli`. That means the downloaded models are immediately usable by:

- **Ollama**
- **LM Studio**
- **llama.cpp**
- **MLX / oMLX**
- **vLLM**
- **Hugging Face `transformers`**

Uninstalling ModelHub leaves all models intact; there is no proprietary format or migration required.

## Core Capabilities

- **Unified model list** — scans LM Studio and the Hugging Face cache on every open
- **Live Hugging Face search** — top text-generation models load by default; type to search in real time
- **Download management** — progress indicator with pause and resume from the menu bar
- **Model management** — copy model ID, reveal in Finder, move to Trash, sort by size or date, filter by name
- **Active model indicator** — a pulsing green dot shows which model is currently loaded in LM Studio

## Architecture and Deployment

The app is written in Swift and runs as a menu-bar agent with no Dock icon. It requires macOS 26+ and Apple Silicon; the binary is approximately 4 MB. Users can install via a `.dmg` download from the GitHub releases page or build from source using Xcode 26+. The project is hosted at `github.com/conscious-engines/modelhub` and released under the MIT license.

## Update: ModelHub 1.3.0

Version 1.3.0 was published on May 18, 2026, and is the current stable release. The repository was created in April 2026 and has seen active development, with the latest push matching the 1.3.0 release. The project's GitHub page lists 73 stars and 7 forks as of late May 2026, signaling early but growing community interest in the tool.

## Features
- Menu-bar access to all local LLMs
- Scans LM Studio and Hugging Face cache directories
- Browse and search Hugging Face text-generation models
- One-click model download with progress, pause, and resume
- Standard Hugging Face cache layout (zero lock-in)
- Reveal model in Finder
- Move model to Trash
- Sort models by size or date
- Filter models by name
- Copy model ID to clipboard
- Active model indicator (pulsing green dot for LM Studio)
- ~4 MB binary, no Dock icon
- Build from source with Xcode 26+

## Integrations
Ollama, LM Studio, llama.cpp, MLX, oMLX, vLLM, Hugging Face, Hugging Face transformers, huggingface-cli

## Platforms
MACOS, CLI

## Pricing
Open Source

## Version
1.3.0

## Links
- Website: https://studio.consciousengines.com/model-hub
- Documentation: https://studio.consciousengines.com/model-hub/changelog
- Repository: https://github.com/conscious-engines/modelhub
- EveryDev.ai: https://www.everydev.ai/tools/modelhub-conscious-engines