How does this differ from OpenClaw? I like that OpenClaw runs on a dedicated machine so it's kind of sandboxed. I think anything you can do in Claude Cowork you can do with OpenClaw + GPT5.6 or another model on the backend. But am I missing some advantage to running this on a daily driver laptop?

Looks cool. I use Hermes Agent currently and love it, but it’s fun to try other tooling too, and openwork isn’t something I’d heard of before. Thanks!

I use omp.sh and git 👋