Fleet

A fleet is the set of other ADI machines you’ve paired with this one. Pairing turns a remote node’s services into ordinary-looking hostnames on your own front door: <service>.<node>.n.adi — a laptop’s nosh dashboard, paired as laptop-b, is nosh.laptop-b.n.adi in your browser, right beside your own <service>.adi names.

Why it exists

Reaching a service on another machine normally means an open inbound port, a VPN, or a tunnel you set up by hand — and remembering which of those you used, on which machine, for which service. Fleet needs none of it: a node only ever dials out, so it can sit behind any NAT with zero inbound firewall rules, and once paired its services just appear under real hostnames on your machine, the same way a local hive service does.

Do it

Panel: /settings/fleet. On the CLI, pairing and grants are grouped under adi-mono mesh (the network layer’s name — see Mesh — though what these commands manage is the fleet):

adi-mono mesh invite                    # mint a one-time invite, on the machine you want to reach
adi-mono mesh join <token>              # spend it, on the machine dialing in
adi-mono mesh fleet                     # list paired nodes: petname, key, grants
adi-mono mesh grant laptop-b http:nosh  # let that node reach one more service here

A phone or a browser tab pairs the other way round — it mints nothing, since nothing can dial into a browser; the node it’s pairing with mints the invite instead, and the tab scans or pastes it.

Two locks, both requiredEnterprise interest

Pairing alone doesn’t open anything. Reaching a node’s service needs both:

• A grant, per node, naming what it may reach (http:app, http:nosh, http:*) — default-deny, so an empty grant list means nobody gets in, not everybody.
• A password, set on the node itself and required over HTTP for every request that arrives through the mesh, including its own control panel. A mesh grant is machine-scoped — anything on your paired laptop could otherwise reach the node — the password is what stops that, and what stops someone else on a shared machine from driving it.

Pairing hands out exactly one default grant, http:app — the node’s own control panel, nothing else. A dashboard isn’t reachable until it’s named.

What’s on the far sideExpert area

A paired node keeps its own full stack — its own hive, its own DNS — so its services resolve locally on it exactly as yours do on you. What crosses the mesh is HTTP, over one pooled connection per peer; see Mesh for the transport itself, and for reaching something that isn’t a hive service at all.