Agentix has reached a v0.1 MVP candidate.

This does not mean the LLM can run the computer unsupervised. It means the safety foundation is in place.

What works

The MVP includes:

  • doctor and status checks
  • safe NixOS verification without switching
  • a high-level goal runner
  • proposal patch creation
  • approval before apply
  • dirty-tree preflight
  • explicit --allow-dirty escape hatch
  • stale patch rejection before approval
  • proposal clean/stale status
  • proposal JSON output
  • stale-only proposal cleanup
  • cleanup and prune JSON output
  • audit tail JSON
  • audit summary JSON
  • installed-command self-test
  • operating guide
  • checkpoint syncing between repos

Core safety contract

Agentix should not run:

sudo
nixos-rebuild switch
direct /etc/nixos mutation

Agentix should stop before activation and tell the human the next safe command.

Useful commands

agentix doctor --path ~/nixos-config
agentix status --path ~/nixos-config
agentix verify --path ~/nixos-config --host nixos
agentix verify --path ~/nixos-config --host nixos --json

Goal runner:

agentix run "add package btop" --path ~/nixos-config --dry-run
agentix run "create module ai with template packages" --path ~/nixos-config --dry-run

Proposal maintenance:

agentix proposals list --path ~/nixos-config
agentix proposals list --path ~/nixos-config --json
agentix proposals clean --path ~/nixos-config --stale
agentix proposals clean --path ~/nixos-config --stale --yes --json

Audit observability:

agentix audit tail --path ~/nixos-config --json
agentix audit summary --path ~/nixos-config --json

Self-test:

agentix self-test

What is intentionally not included

Agentix v0.1 does not include:

  • automatic live system switching
  • direct sudo operations
  • broad natural-language config editing
  • model/provider integration
  • Home Manager support
  • unsupervised live machine operation

Those are future phases. The MVP is about making the safety loop boring and reliable.