Linux

Local AI meeting notes on Linux

LokalBot records both sides of a call without a bot, transcribes and summarizes it with models on your computer, and keeps every meeting searchable. Here is what works on Linux, what depends on your desktop, and how to install it.

What works on Linux

LokalBot for Linux is a native app written in Rust with GPUI, the UI framework behind the Zed editor. It has the same pages as the Mac app (Today, Meetings, Timeline, People, Projects, Ask, and Settings) and reads the same library format, so a library recorded on a Mac opens on Linux. It's an alpha.

Your desktop matters

Meetings, transcription, notes, search, and Ask work on any desktop. A few features depend on what your desktop lets apps do:

FeatureX11Wayland
Day memory (app and window activity)YesHyprland, Sway, and KDE; GNOME with the Window Calls extension
Screen context (visible text, optional encrypted screenshots)YesSway and Hyprland with grim; other desktops through the screenshot portal when it can capture just the active window
DictationYesBind lokalbot dictate to a shortcut in your compositor
Autocomplete in other appsGTK apps; Qt, Chromium, Electron, and Firefox once system accessibility is onOnly in the practice editor in Settings
Tray iconYes; GNOME needs the AppIndicator extension

On Omarchy and other Hyprland desktops, grim, wtype, wl-clipboard, and PipeWire are already there, and the dictation pill and Quick Recall float instead of tiling.

Install and stay updated

All of them are on the latest release. LokalBot needs x86_64 with glibc 2.39 or later (Ubuntu 24.04, Debian 13, Fedora 40, or newer), PulseAudio or PipeWire, and a Vulkan or OpenGL driver; Mesa is fine. See the system requirements.

Models and hardware

On first launch LokalBot asks what to remember and downloads the models you choose: about 4 GB for the default transcription, notes, and search models. llama.cpp runs on any Vulkan GPU or on the CPU. On an NVIDIA GPU you can choose CUDA (experimental) in Settings → Models → llama.cpp; in testing, CUDA read a long prompt into Qwen3.5 4B 44% faster than Vulkan on an RTX 3060. Without a capable GPU, notes on the CPU are slow, so you can point notes and Ask at Ollama on another computer or at a server you approve.

What's not there yet

FAQ

Questions about this guide

Does LokalBot work on Wayland?

Yes, with limits. Meetings, transcription, notes, search, and Ask work on any desktop. Day memory works on Hyprland, Sway, KDE, and GNOME with the Window Calls extension; screen context needs grim on Sway and Hyprland or the desktop's screenshot portal elsewhere; dictation needs a compositor shortcut bound to lokalbot dictate; and autocomplete in other apps needs X11.

Do I need an NVIDIA GPU?

No. The built-in models run on any Vulkan GPU, including AMD and Intel graphics, or on the CPU. CUDA is an experimental option for NVIDIA GPUs.

Can I move my library between Linux and a Mac?

Yes. Both apps read the same library format, and meetings recorded on a Mac play and transcribe on Linux. On Linux the library lives in ~/.local/share/LokalBot. LokalBot doesn't sync it between computers for you.

Try it locally

Meeting notes that stay on your computer by default.

Free, open source, and no account. For Mac, with Windows and Linux in alpha.

Download for macOS

Apple Silicon · macOS 15+ · or with Homebrew: brew install --cask stevyhacker/tap/lokalbot

macOS · Windows · Linux (alpha) · requirements