Local LLMs on Linux.

AppImage or tar.gz. CUDA on NVIDIA. The simplest local AI you can self-host.

Linux is where local AI workflows live for a lot of developers , GPU drivers are cleaner, package management is simpler, and there's no telemetry layer sitting between you and the kernel. tailor. ships as an AppImage that runs on any modern distro, with the same agentic features, document chat, and local API as the Mac and Windows builds.

Distributions tested

Ubuntu 22.04 and 24.04 LTS. Debian 12 (Bookworm). Fedora 38, 39, 40. Pop!_OS 22.04. Arch (current). Anything with glibc 2.34+ and a recent X11 or Wayland session should work , the AppImage bundles its own dependencies otherwise.

GPU acceleration

NVIDIA: CUDA picked up automatically if you have a recent driver. CPU-only fallback always available and works on any 64-bit Linux. ROCm and Vulkan acceleration for AMD/Intel GPUs are on the roadmap , for now those run on the CPU path.

Installation

Two distribution formats: a tar.gz with desktop file and icons, or an AppImage. Download the .AppImage, chmod +x, run. The app self-contains its dependencies. Optional: integrate into your menu with appimaged or gear-lever. Native .deb and .rpm packages are on the roadmap.

Questions

Does the AppImage need FUSE?

Yes , most distros include it by default. On systems without it, run with --appimage-extract-and-run as a workaround.

Can I run tailor. on a headless server?

Not yet , tailor. is a desktop app today and the GUI is required to drive it. A dedicated headless mode for running the inference server on a LAN box without the desktop window is on the roadmap.

Is there a Snap, Flatpak, .deb, or .rpm?

Not currently. Today the supported Linux distributions are an AppImage and a tar.gz with desktop file and icons. Native .deb / .rpm / Flatpak builds are on the roadmap.

Try tailor. free for 7 days.

Full access. No credit card required. Mac, Windows, and Linux.

Start free trial → See pricing