Skip to content

Install & run Understand fully offline

Understand installs and runs on a machine with no internet access — analysis, metrics, CodeCheck, reports, and graphs all run locally. Each platform has a kit that needs no connection on the target machine, and this page orchestrates the install and licensing steps for a disconnected environment; it doesn't repeat them.

The only things that need a network are optional

Installation and code analysis never need the internet. The features that can reach out — online licensing, update checks, analytics, AI model download, and cloud AI — are each optional and avoidable. See What leaves the machine for the exact list, and Licensing without internet for the offline licensing flow.

1. Get Understand onto the machine

  • Windows — the standard installer is a small web installer that downloads components as it runs, so it isn't the right vehicle for an air-gapped machine. Instead, build a portable .zip on a connected machine — components and local AI model included — and carry that over: see Create an offline install kit for Windows.
  • macOS and Linux — the .dmg and .tgz kits are self-contained downloads; transfer one by your approved method (USB, internal artifact repo, etc.) and install as normal — see Install Understand.

For servers, CI runners, and containers with no display, use the CLI-only kit — see the Headless / CI / Docker tab on the install page. The CLI kit ships und, uperl, upython, userver, and uvalid and needs no GUI.

Graphs need X11 on headless Linux

Report and metric generation work fully headless, but generating graph images (GUI or API) requires X11 libraries on the machine. This is a local rendering dependency, not a network one.

2. License it without internet

An offline machine can't reach the online licensing service, so use one of the disconnected licensing methods:

Both work from the GUI or from the und command line, so they fit headless and scripted installs. Full walkthrough: Licensing without internet.

3. Keep it disconnected

Once installed and licensed offline, Understand does not need to reach the internet again to analyze code. To be certain nothing reaches out, review and disable the optional network features:

  • Turn off analytics and (on Windows) automatic crash reporting.
  • Stay in offline/node-lock licensing mode rather than online mode.
  • Keep AI on the bundled local model (or disable AI entirely).

Each of these is covered, with the exact settings and source-verified endpoints, in What leaves the machine and Local AI with no cloud calls.

Docker / containers

The official Docker image is built from the Linux CLI tarball and runs headless. In an air-gapped network, pull or load the image into your internal registry first, then run it there. See the Docker notes on the install page (Headless / CI / Docker tab).