Summary
Cate already lets one terminal drive another through the cate CLI (cate terminal read / type / press, plus panel verbs). That orchestration works well — but it is invisible on the canvas. When several agents coordinate (orchestrator → workers, a review loop, research → build), nothing on the surface shows which panel talks to which.
A visual connector — an arrow drawn from one panel to another — would make those relationships legible, and could optionally carry flow semantics.
Use case
With --dangerously-skip-permissions agents fanned out across the canvas, it gets hard to remember the topology:
- panel A researches, hands findings to panel B which builds
- panel A (orchestrator) delegates to B and C, collects results
- A writes code, B reviews, back and forth
Today the only way to convey this is renaming panels ("1·backend", "2·frontend") or a separate note. An arrow between two panels would show it at a glance and survive in the saved .cate/ layout.
Proposed behaviour
Phase 1 — purely visual (low effort, high value):
- Drag from a panel edge/anchor to another panel to create an arrow.
- Arrows live in canvas-space (pan/zoom with the world), snap to panel anchor points, are curveable and reconnectable, deletable, and persist in the workspace layout.
- Optional label on the arrow ("findings", "review", "deploy").
This alone documents the mental model of a multi-agent session — no execution semantics required.
Phase 2 — optional flow semantics (later):
- An arrow from agent A to agent B could carry a trigger, e.g. "when A finishes its turn, pass its last output to B as a prompt." This turns a diagram into a runnable pipeline. A comparable model exists in other canvas-agent tools ("logic arrows"), but Phase 1 is valuable on its own and much cheaper.
Why it fits Cate
The canvas is already the core metaphor, panels already have stable ids and anchor geometry, and the layout already persists per-workspace in .cate/. Connectors are a natural extension of the spatial model, and they pair directly with the existing cate CLI orchestration that has no visual counterpart today.
Environment
- Cate 1.5.3 (Homebrew cask), macOS 15, Apple Silicon
Summary
Cate already lets one terminal drive another through the
cateCLI (cate terminal read/type/press, pluspanelverbs). That orchestration works well — but it is invisible on the canvas. When several agents coordinate (orchestrator → workers, a review loop, research → build), nothing on the surface shows which panel talks to which.A visual connector — an arrow drawn from one panel to another — would make those relationships legible, and could optionally carry flow semantics.
Use case
With
--dangerously-skip-permissionsagents fanned out across the canvas, it gets hard to remember the topology:Today the only way to convey this is renaming panels ("1·backend", "2·frontend") or a separate note. An arrow between two panels would show it at a glance and survive in the saved
.cate/layout.Proposed behaviour
Phase 1 — purely visual (low effort, high value):
This alone documents the mental model of a multi-agent session — no execution semantics required.
Phase 2 — optional flow semantics (later):
Why it fits Cate
The canvas is already the core metaphor, panels already have stable ids and anchor geometry, and the layout already persists per-workspace in
.cate/. Connectors are a natural extension of the spatial model, and they pair directly with the existingcateCLI orchestration that has no visual counterpart today.Environment