Repository navigation
Safety and agent-first Linux distro at the edge #2
zhouwg
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Safety and agent-first Linux distro at the edge
Inspired by Omarchy ("Convention over Configuration") and Omarchy's desktop shell, and drawing inspiration from local AI projects such as llama.cpp, there is growing interest in running AI agents fully on-device, without mandatory cloud dependency. Edge hardware is capable of hosting autonomous agents, yet most existing Linux distributions are not built with agent safety as a first-class design goal.
The past months brought many interesting discussions with developers exploring local AI workflows. It became clear that we need a minimal, auditable Linux distro designed specifically for AI agents to operate safely on edge hardware.
The AI agent space has plenty of exciting directions. Many ideas are experimental, not all will end up useful, but building them locally is deeply rewarding. From agent-assisted system maintenance, automated configuration refactoring, to running local LLM reasoning on consumer x86_64 hardware, the possibilities are broad. Shifting agent workloads away from centralized cloud and onto the user’s own machine is what motivates this work.
Naturally, I’ve started building Kan-Linux around these goals. These are the core principles so far:
I hope this resonates with developers interested in secure local agent computing. I welcome your ideas, patches and feedback on what an agent-native edge Linux distro should become.
All reactions