The canvas
Where a tron is drawn. Agents are objects you pick up, wire and open.
Moving around
Drag to pan, Ctrl and scroll to zoom, and 0 to reset the view.
Adding an agent
An agent takes a model from the catalogue: only the models switched on there can be picked. Its name is its label, and the label is an address: the routes of the site reach an agent as /trons/:id/nuclei/:agent, so a rename moves it. From code use its id (nucleus_id in the Tron API), which a rename does not change.
Wiring
A wire between two agents is an action (the first delegates a task to the second and waits for its answer, shown in the chat as a hand-off) or a message (sent, without waiting). A wire from a memory to an agent gives that agent the memory, read only or read and write.
Memory nodes
One memory can be wired to several agents, each with its own access. The five kinds are buffer, summary + buffer, wiki, entity and RAG; any of them can be shared or per session. The details, and the runtime a per-session wiki, entity or RAG memory needs, are in Memory.
Saving
There is no save button. An edit is a request, and the only place that reports whether it landed is the write status in the header.