Very nice. Can you elaborate why you chose mTLS authentication vs. SSH as a mean to deploy the policies.
Regarding the docs, the diagrams should really be replaced with something more familiar and readable such as Mermaid (currently it looks like a mixture of ad-hoc ASCII charts).
The general documentation is something that definitely needs a lot of improvement, but it's currently under active development.
Policies are delivered over a persistent gRPC stream, secured with mTLS, where each node gets it's own cert from Bor's built-in CA at enrollment - so there's no SSH key sprawl and no credentials on the server that could log into machines. Since agents connect outbound to the server, it works through NAT and firewalls without opening any inbound ports on desktops, and policy changes propagate in seconds over the already-open stream. SHH-push would have meant maintaining an inventory of searchable hosts and a server that can shell into the whole fleet - a much bigger attach surface for less capability.
Some orgs might not want ssh running on user laptops/workstations, so I think that it's a plus to not have that requirement.