LLamaDesk
Private by design
LLamaDesk only talks to the Ollama address you set. It loads no external fonts, scripts, analytics, or tracking, and your chats are stored on your machine.
Chat with your local models on your desktop.
LLamaDesk is a fast, minimal desktop app for Ollama. Your prompts, files, and chat history never leave your machine.
Everything you need to talk to a local model
A small native app that stays out of your way. It connects straight to Ollama and nothing else.
Streaming you can steer
Watch replies arrive token by token, see the speed, and stop a response mid-sentence.
Stop42.8 tok/s
Send images to vision models
Attach, drag and drop, or paste. Images are resized and converted before they're sent. Works with models like llama3.2-vision, llava, and gemma3.
AttachDrag & dropPaste
Text and code files
Drop in a file and ask about it. Files up to 400 KB are added to your message.
.txt.md.csv.json.py.js
Searchable history
Chats are saved locally. Search them, or delete any one you don't want to keep.
Markdown and reasoning
Code blocks, tables, and lists render cleanly. Reasoning stays tucked away until you want it.
Reasoning
The user wants to keep the first occurrence of each item, so a set can track what has been seen.
Model picker
Lists every model installed in Ollama. Switch from the input box at any time.
Copy and regenerate
Copy any reply, or regenerate the latest one with a click.
Settings that matter
Set a system prompt, temperature, the Ollama address, and a dark, light, or system theme.
Native and light
Built with Tauri 2 for a small binary and low memory use.
Keyboard friendly
Enter to send, Shift+Enter for a new line.
Responsive layout
The sidebar collapses so the window works at any size.
Download LLamaDesk
Installers are published on GitHub Releases for every version.
Windows
Recommended for youSetup installer or MSI package.
Linux
Recommended for youAppImage for any distribution, or a .deb for Debian and Ubuntu.
macOS builds are on the roadmap but need an Apple Developer certificate. Want to build it yourself? See build from source.
Install Ollama
LLamaDesk needs Ollama installed and running on your computer.
Pull a model
Download at least one model from your terminal.
ollama pull llama3.2Open LLamaDesk
It lists every model you've installed. Pick one and start chatting. For images, use a vision model.
Questions
The short answers.
Is LLamaDesk free?
Yes. It's open source under the MIT license, so you can use it, read the code, and contribute.
Do I need Ollama?
Yes. LLamaDesk is a desktop interface for Ollama, so Ollama needs to be installed and running, with at least one model pulled.
Does my data leave my computer?
No. LLamaDesk only connects to the Ollama address you configure, which is http://localhost:11434 by default. Chats and settings are stored locally, and the app includes no analytics or tracking.
Which platforms are supported?
Windows and Linux today. macOS is planned, but it requires an Apple Developer certificate.
Can I attach PDFs or Office documents?
Not yet. Text and code files work, and images work with vision models. For PDFs, convert them to text first, for example with pdftotext. In-app PDF extraction is on the roadmap.
Why can't it reach Ollama?
Check that Ollama is running and that the address in Settings is correct. You can test it with curl http://localhost:11434/api/tags. If no models appear, pull one and press Reload models in Settings.
Is this made by Ollama?
No. LLamaDesk is an independent project and isn't affiliated with or endorsed by Ollama or the Tauri project.