OpenAltFinder
Hob

Hob

Free, MIT-licensed desktop AI assistant that runs open-weight models like Gemma and gpt-oss locally on Mac, Windows, and Linux.

Open source alternative to:

screenshot.png

Hob is a desktop AI assistant that runs open-weight language models entirely on your own computer. It is built by Frigade as a free, MIT-licensed alternative to Claude and ChatGPT, aimed at people who want the same kind of file-and-tool assistance without sending their documents to a cloud service.

Hob inspects your hardware and downloads the largest model your machine can run well, then handles the setup itself. The curated lineup includes Google Gemma 4 variants (E4B Compact on 8 GB Macs, E4B on 16 GB, 26B from 24 GB) and OpenAI gpt-oss 120B for workstations, plus a small Qwen3-VL model for images. It works on files, folders and projects with their own instructions, reads and fills PDFs, analyzes spreadsheets, writes and runs Python in a sandbox, draws interactive charts, and searches the web through a metasearch plugin that runs locally. File changes require approval and can be undone, and the app is translated into 32 languages.

Like Claude and ChatGPT, Hob is built for agent-style work rather than single questions: it plans steps, calls tools, and returns finished documents you can open and edit. The difference is privacy and cost. Chats, files and the model stay on the machine, there is no account or subscription, and web access can be switched off for fully offline use. The trade-off is hardware: you need at least 8 GB of RAM, and output quality on smaller machines depends on the lighter model Hob selects. Hob ships as a native app for Apple Silicon Macs, Windows and Linux rather than a self-hosted web service, and its skills, Python tools and MCP connectors are extensible through plugins.

Frequently Asked Questions

Which platforms does Hob run on?

Hob is a desktop app for Apple Silicon Macs, Windows PCs, and Linux (AppImage). There is no web app, mobile app, or official CLI; you download an installer from the GitHub releases page or hob.chat.

What are the system requirements for Hob?

You need at least 8 GB of RAM and a reasonably modern 64-bit CPU; Macs must be Apple Silicon. 16 GB is comfortable for everyday tasks and 32 GB or more lets Hob run its largest models. Plan for roughly 10 to 25 GB of free disk space for the app and models, or about 75 GB if you use the largest model. Models run through llama.cpp with Metal on Mac and CUDA or Vulkan elsewhere.

Can Hob work fully offline?

Yes. Turn off Web access in Settings under Privacy and Hob works offline; chats and file work never leave the computer. Internet is needed to download the app and its models, to install updates, and when you use web search, connectors, or install Python packages from PyPI.

Which models does Hob use, and can I switch them?

Hob picks a model based on your memory and can be changed in Settings. The list includes Google Gemma 4 (E4B Compact on 8 GB Macs, E4B on 16 GB, 26B from 24 GB) and OpenAI gpt-oss 120B for workstations. A separate small Qwen3-VL model is downloaded for reading images and screenshots.

How do Hob plugins and connectors work?

A plugin is a folder containing a hob-plugin.json file. It can add skills in the Agent Skills format, Python tools, and connectors to any MCP server, including .mcpb bundles. Hob's own web search, code execution, and PDF features are implemented as plugins, and you can ask Hob to save a repeated task as a new skill.

OpenAltFinder Score
67/100
Project Health (63%)
47/100
License (25%)
100/100
Recency (13%)
100/100
How scoring works
Project Details
Platforms
Desktop
License
MIT
Self hostable
No
Repository details
Created
10/14/2026
Stars
148
Forks
4
Last commit
10/27/2026
View Repository

Get the latest open source alternatives delivered weekly.
Discover tools that give you freedom from proprietary software.

Similar open source alternatives