Node properties modal
You've double-clicked a node and a large panel has opened over the canvas. This page explains what's in it, how saving works, and the few nodes whose panel looks a little different.
It's for builders configuring nodes for the first time. If you haven't placed a node yet, start with Canvas Basics.

Opening it
Double-click a node, or hover it and choose More actions → Open… (the Execute step button above a node opens the same panel). The header shows the node's type and name, a save indicator, a Docs ↗ link to the help page for that node, and ✕ to close. Escape also closes it.
The three panes
| Pane | Where | What it shows |
|---|---|---|
| Input | Left | Available variables — everything this node can use, grouped by where it comes from (the trigger, the customer, each earlier node). Search them at the top. |
| Config / Settings / Notes | Centre | The node's settings. |
| Output | Right | Variables this node emits — what later nodes can use. Variables marked if produced only appear when the node actually creates them. |
A Phone Call trigger also offers the call's own details — caller (the number that placed the call), called_number, direction (inbound or outbound) and call_purpose (why an outbound call was placed). An AI Agent on a voice call also produces call_summary — a summary of the call, available to steps after the agent once the call ends.
Next to each variable, ↵ inserts it into the field you're editing and the copy icon copies its path. Parts of an AI Agent (Personality, AI Model…) see the AI Agent's variables. A trigger's Input pane just says Workflow entry, because nothing runs before it; nodes that don't produce variables say so in the Output pane (No workflow variables emitted, or Configuration only for an agent part).
The centre tabs
| Tab | What's there |
|---|---|
| Config | The node's own settings — different for every node type. |
| Settings | Node Label (the name on the canvas) and Description; for nodes that can be retried, Max Attempts, Backoff (ms) between attempts, and Timeout (ms). |
| Notes | Free-text notes for your team. They're stored on the node for documentation only and never affect how it runs. |
A few nodes have something extra:
- Web Chat trigger — the Config tab has Configure, Appearance and Install views, a Live preview that runs your draft in a real widget, and a Focus button that hides the side panes. See Trigger Nodes.
- Schedule and Manual Test triggers — a Run now / Execute button at the top of Config runs the agent's draft once; the run appears in Runs. See Testing before you go live.
- Operator — an extra Operator app tab for setting up the operator app.
Fixing errors
If the node has something that stops the agent from being activated, a red box at the top of the modal lists every problem on that node — "Fix these before you can publish" — so you can fix them all in one visit. Clicking an error in the canvas's validation banner opens the node it belongs to.
Saving
There's no Save button. Changes save automatically about a second and a half after you stop typing, and the header shows Saving… then Saved. While the modal is open, canvas shortcuts (Delete, undo, copy) are switched off, so typing in a field never changes the canvas behind it.
If the agent is Inactive, your change takes effect when you Activate it. If it's already Active, the change is saved as unpublished changes — customers keep the live version until you click Publish changes in the toolbar. See Publishing changes to an active agent.
Example. Asha opens the Personality node of Acme Support's live agent, changes the greeting, pauses, and sees Saved. She presses Escape. The toolbar now shows Unpublished changes; she tries the new greeting in the Web Chat live preview, then clicks Publish changes — only then do customers get it.
Looking at a node inside a past run
When you open a node from the Runs view, the modal shows what that node did in the selected run instead of its settings contract:
- a Read-only badge in the header — the configuration is shown so you can see what produced the run, but nothing can be edited;
- the Input and Output panes show that node's actual data in the run (Output · this run, with any error, the input and the output);
- an action the AI Agent called (directly or as a step of a tool chain) shows its most recent call in that run;
- a node that didn't run in that execution says so.
See Runs & Executions.
Next steps
- Canvas Basics — adding, connecting and activating.
- AI Agent Node — every setting on the AI Agent.
- Trigger Nodes — trigger settings, including Web Chat's Install view.