Skip to content

MESA KASM Ubuntu Desktop

Repo: idss-mesa/kasm · Image: harbor.cyverse.org/vice/mesa-kasm:latest (GPU: :gpu) · In the portal: Applications → MESA Apps → MESA KASM Ubuntu Desktop

A complete Ubuntu 24.04 desktop (XFCE) in a browser tab, streamed with KasmVNC1. Use it for software that needs a graphical interface — desktop GIS, 3D viewers, or anything you install yourself — next to the MESA AI coding agents and CyVerse Data Store tools.

What's inside

Category Tools
Desktop XFCE with a terminal, Chrome, Firefox, VS Code, and the rest of the Kasm desktop app set
AI agents and MCP Claude Code, Codex, OpenCode, Antigravity, Claude Code Router; the irods, mesa, formation, and filesystem MCP servers — see AI agents in the MESA apps
CyVerse data GoCommands, iRODS configuration, S3/OSN mounts, AWS CLI
Developer tools GitHub CLI, Git Credential Manager, Go 1.25, Node.js 22

Start it

  1. In the MESA Portal, open Applications → MESA Apps.
  2. On MESA KASM Ubuntu Desktop, click Instant Launch or Launch with Options. See Starting applications.
  3. The desktop opens in a new tab. Your Data Store is in ~/data-store.

You can also start it from the Discovery Environment by searching for MESA KASM Ubuntu Desktop.

First steps

  1. Open a terminal from the desktop.
  2. Run cyverse-login to give the tools and agents your CyVerse access.
  3. Optionally run aiverde-setup to connect AI Verde models.
  4. Start an agent: claude, codex, opencode, or agy.

See AI agents in the MESA apps. Save files you want to keep under ~/data-store; the rest of the desktop is deleted when the analysis ends.

GPU build

harbor.cyverse.org/vice/mesa-kasm:gpu runs the desktop on an NVIDIA GPU. It adds:

Adds Details
Desktop graphics OpenGL on the GPU for the whole desktop (browsers, VS Code, Qt apps, and 3D apps you install such as Blender); mesa-gl-run <app> (VirtualGL) for heavy 3D; glxinfo, vulkaninfo, glmark2; GPU Monitor, GPU Check, and GLX Spheres in the menu
PyTorch PyTorch 2.14 (CUDA 12.6) in the conda environment pytorch: run conda activate pytorch
Local LLMs An Ollama server on the GPU — see Local models on a GPU
GPU tools mesa-gpu-check, nvtop, nvitop

If a 3D app is slow or draws incorrectly, start it with mesa-gl-run <app> instead, or with __GLX_VENDOR_LIBRARY_NAME=mesa <app> to fall back to software rendering for that app only.

Run it on your own computer

docker run --rm --shm-size=512m -p 6901:6901 -e IPLANT_USER=$USER harbor.cyverse.org/vice/mesa-kasm:latest

Open http://localhost:6901. Outside CyVerse there is no password, so publish the port only on your own machine. The image is built for linux/amd64.

  1. MESA KASM Ubuntu Desktop README, https://github.com/idss-mesa/kasm. ↩

Machine-readable versions of this page: Markdown twin · raw source on GitHub · llms.txt · llms-full.txt (whole site). See For AI agents.