Hob - Open-source, on device alternative to Claude and ChatGPT
by•
Hob is a free, open-source AI app for Mac, Windows and Linux that works with your files the way ChatGPT and Claude do, except the AI runs on your own computer. It picks the best open model your computer can run and sets it up for you. Hob can read and edit files in folders you choose, turn spreadsheets into charts, fill in PDFs, look at photos and search the web. Your chats and files stay on your computer, and it works offline.
Replies
Hey Product Hunt! I'm Christian and Hob is a new open-source project I've been noodling on for the last couple of weeks.
Hob is a free desktop app for Mac, Windows and Linux that works on your files the way the Claude and ChatGPT apps do, except the AI runs on your own computer. Your chats and files never leave your computer, and there's no account or subscription.
Some of what it can do:
Picks the best open model for your computer and sets everything up for you
Reads and edits files in folders you choose (it asks before changing anything, and you can undo)
Turns spreadsheets into charts, fills in PDFs, and works with Word, Excel and PowerPoint files
Looks at photos and screenshots
Searches the web, or you can turn that off and stay fully offline
Connects to your other apps through MCP
Before building this I tried every local AI app I could find, but it seemed each one had different problems. None of them could do what the ChatGPT and Claude apps do with your files, e.g. read a folder, look at an image, draw a chart. Most of them would paste a text file into a small context window, which failed as soon as the file was larger than a few KB. The ones that could do more seemed to default to hosted models, which defeats the entire point of keeping things private, or push you towards a paid plan.
Obviously local models are weaker than the newest frontier models. But for most tasks for most normal people, I'd argue you rarely need the newest frontier model. From my own testing, computers with 16+ GB of memory do really well, and with more memory you get very close to what you'd get from the hosted apps.
Would love to hear what you think and how it runs on your computer!
Undo matters when the agent is editing files you already use. I like that it's part of the workflow from the start.
@oleksii_sekundant thank you - yes, don't want my mom to delete all her photos by accident :)
@nicholas_zakas 8gb RAM alone is enough to run a decent Gemma model. Highly recommend testing it out!
@cmathies Yeah, that's what I had tried. I ended up needing to run it on the CPU and then could only get E2B running at anything close to usable. I can give Hob a try and see if it behaves any differently for me.
The file-editing part is what stood out to me — asking before changing anything and having undo built in, especially for spreadsheets and PDFs where a bad edit is hard to spot. When Hob is working across a folder you chose, does it show you a diff or preview before it writes, and for a large file does it read just the relevant part or load the whole thing? I'm curious how that stays reliable once a file is bigger than a few KB.
Watching the runtime-selection logic in the comments - that's the real win here. Most local AI apps default to cloud or force you through setup hell. Auto-detecting hardware (CPU/GPU/RAM) and picking an appropriate Ollama model, then *asking before any file changes* - that's thinking about how people actually use their machines. Looking forward to seeing how this scales with MCP integrations.