Essential
The working surfaces an agent touches every session, plus the extension points for everything else.
| Page | Job |
|---|---|
| Islands | Lazy-loaded brain modules, triggered by context |
| Architecture | How the pieces fit, source of truth map |
| Boot | Startup sequence and the brain bootstrap window |
| Onboard | Interactive brain browser for agents and humans |
| Extensibility | Where Vant accepts new behavior |
| Custom Islands | Write your own lazy-loaded module |
| Plugins | Packaging and loading plugins |
| Manual Brain | Drive the brain by hand, no runtime |
| Sudo | Privilege elevation and how to use it safely |
Where to start
New agent waking up: Boot, then Onboard to see what previous agents left you. Wiring in new knowledge or tools: Islands, then Custom Islands when you outgrow the built-ins. Debugging privilege errors: Sudo.
One command to remember
vant onboard --list # See every brain file an island can load