Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUs

Your AI just spun up cloud PCs and GPUs — devs cheer, skeptics squint

TLDR: A new CLI skill, cloudrouter, lets coding AIs quickly create cloud computers, run commands, automate a browser, and add powerful GPUs. The crowd split between hype for one‑click convenience, a bug report over a surprise login prompt, and debate over whether it’s better than just using existing cloud tools.

Cloudrouter crashed onto Show HN promising a one-liner that lets coding bots like Claude Code summon disposable cloud computers, open VS Code in the browser, run terminals, even drive a built‑in Chrome to click buttons and take screenshots — with optional “superchip” GPUs for heavy lifting. The demo had fans yelling “instant playground!”, with one early tester calling it “Awesome.”

Then the plot twist: a user hit a surprise password prompt while trying it outside an agent and filed a GitHub issue. Cue the classic HN energy: “Cool idea, but rough edges.” Meanwhile, the practicality police showed up asking why we need a new tool at all — why not just tell the agent to use AWS/Azure/Google’s existing command tools? That kicked off a convenience vs. complexity scuffle: fans say a single skill keeps bots from fumbling 20 different cloud knobs; skeptics call it a shiny wrapper.

And of course, the dystopia jokes wrote themselves. One commenter quipped we’re “one step closer to a model bootstrapping itself.” Translation: your bot spins up a machine, installs code, controls a browser, maybe grabs a GPU — and now everyone’s making Skynet memes. It’s part product launch, part philosophy debate, with a side of “my laptop is now a browser tab… plus a rental bill.”

Key Points

  • Cloudrouter can be installed as an agent skill or standalone CLI to create and manage cloud sandboxes.
  • It supports multiple access modes: VS Code in browser, terminal (PTY), SSH commands, VNC desktop, and Jupyter Lab.
  • Full Chrome CDP integration enables browser automation, including element interaction and screenshots from the CLI.
  • File transfer includes upload/download with watch mode for automatic re-uploads and built-in Docker support.
  • GPU instances are available via flags, with options like T4, L4, A10G, L40S, and H100, including multi-GPU configurations.

Hottest takes

"Awesome demo!!!" — wpan25
"What stops just mentioning AWS/Azure/GCP CLI tools to agents?" — lostmsu
"one step closer to a model with it's own weights file can bootstrap and run itself" — pixl97
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.