Subagents & loops

When an agent branches out — spawning subagents or setting up recurring work — nodeterm draws that structure on the canvas so you can see it happening, not just infer it from scrollback.

Subagent fan-out

When Claude launches subagents, nodeterm shows each one as an ephemeral card connected by an edge to the parent agent node. Each card shows:

the subagent’s type and task,
whether it’s working or done,
its duration, tokens, and tool-use count.

Click a card to expand a live transcript of what that subagent is doing — assistant text, tool calls, and results, streamed as they happen.

These cards are display-only and never persisted: they live outside your saved layout, so they’re not part of undo, dirty state, or the project file. They’re cleared automatically on the next turn, when the session ends, or when you close the node.

Loop, schedule & cron cards

Recurring work gets its own card, connected to the parent agent node. nodeterm recognizes it from the tools the agent invokes — so it works even when you phrase the request in natural language rather than typing the slash command:

/loop — repeats within the current session. The card counts the turns as they go, and the loop dies with its session.
/schedule and /cron — run in the background on a schedule. These cards outlive turns, sessions, and even app restarts, so a scheduled job you set up yesterday is still tracked today.

A schedule or cron card is cleared when the underlying job is deleted, or you can dismiss the card itself from its header.