1. Install and launch

Releases ship as signed binaries for macOS, Linux, and Windows. There is no toolchain to install and no build step — download, extract, run.

All downloads on this page come from the latest release page.

Match the tarball to your CPU

amd64 is Intel and AMD. arm64 is Jetson, Raspberry Pi, and Apple-Silicon Linux. Running the wrong one fails with ld-linux-aarch64.so.1: No such file or a similar loader error — that message means wrong architecture, not a broken download.

macOS (Apple Silicon)

Download infernode-*-macos-arm64.dmg, open it, drag InferNode to Applications, and launch it.

Linux (x86_64 or ARM64)

Download infernode-*-linux-amd64-gui.tar.gz — or -arm64- for Jetson, Raspberry Pi, and Apple-Silicon Linux. SDL3 is bundled, so there is nothing else to install.

tar xzf infernode-*-linux-amd64-gui.tar.gz
cd infernode-*-linux-amd64-gui
./infernode

Optionally, add an app-menu icon and put infernode on your $PATH:

./setup-desktop.sh

./setup-desktop.sh --no-path installs the icon only, --no-icon the PATH wrapper only, and --uninstall removes both.

Windows (x86_64)

Download infernode-*-windows-amd64-gui.zip and extract it. Then, in order:

1. Double-click setup-windows.bat
2. Double-click InferNode.exe

Run the .bat first — this step is not optional

Windows tags browser-downloaded zips with the Mark of the Web and propagates that tag to every extracted file, so SmartScreen silently blocks InferNode.exe. .bat files are exempt from that gate, which is why setup-windows.bat can run at all — clearing the tag from the bundle is the first thing it does. It also configures an LLM backend while it is there.

Windows code signing is provided by the SignPath Foundation, a non-profit that signs open-source releases with certificates issued by SSL.com.

Headless, container, or server

No GUI, no display, no problem:

docker run -it ghcr.io/infernode-os/infernode:latest

The image is multi-arch (amd64 and arm64) and carries SLSA build provenance. For a bare tarball instead, take infernode-*-linux-amd64.tar.gz (no -gui) and run ./infernode-headless.

You will land at the Inferno shell prompt, ;, rather than in a window. Every later step marks the terminal path where it differs.

Verify

With the GUI, you should see a window in three zones, with a welcome document already loaded:

• Conversation — where you talk to the agent
• Presentation — where documents, apps, and output appear
• Context — tool toggles, granted paths, and what the agent knows

The welcome document is /lib/veltro/welcome.md, displayed in the presentation zone on first launch.

If instead Veltro greets you with a dialogue titled LLM Setup, that is expected on a fresh install — it means no model is configured yet, which is exactly what the next step is for.

Headless, you should get a prompt and a version:

; cat /dev/sysctl
Fourth Edition (20120928)

That is InferNode reporting its own kernel version by way of a file, which is the entire idea of the system and the subject of step 4.

Lucia

The three-zone window is Lucia, InferNode’s GUI. It is not a terminal emulator with panes — each zone is a separate surface that programs and agents write to through the filesystem, which is why an agent can put a document in front of you without any GUI toolkit being involved.

Checking the download (optional)

Every release asset is published with a cosign bundle (.pem and .sig) and a signed SHA256SUMS.txt. If you verify signatures as a matter of habit, they are on the release page next to the binaries.

Troubleshooting

ld-linux-aarch64.so.1: No such file — wrong architecture. Get the other tarball.

Windows: nothing happens when you double-click InferNode.exe — you skipped setup-windows.bat. SmartScreen is blocking it silently. Run the .bat, then try again.

The window opens and closes immediately — run the binary from a terminal instead of a file manager so you can read the error it prints.

---

Next: Connect a model — Veltro needs something to think with before it can do anything.