Skip to main content
Running nodes one at a time is fine for quick experiments. The real power of Runchat is chaining nodes together so the output of one flows straight into the next. There are three ways to connect nodes: drawing edges, inline references with @, and showing parameters as inputs.
Two nodes connected by a coloured edge

Drawing an edge

To connect two nodes, drag from an output handle on the right side of one node to an input handle on the left side of another. You can also drop the wire straight onto an input row. Dropping it onto a prompt field inserts a reference at that spot instead. From then on, whenever you run the downstream node, Runchat automatically grabs the latest result from the upstream one and feeds it in. Runchat checks compatibility as you connect. If you try to plug a video into a parameter that expects text, it won’t let you, and it’ll show why.

Dropping an edge on the canvas

Drag from an output and release on empty canvas to open the node picker at that spot. Pick a node or model and it’s created there, already connected to the output you dragged.

Edge colours

Edges are colour-coded by the type of data flowing through them. You don’t need to memorise these, at a glance the colour tells you what kind of data is moving: For the full list of types and how Runchat casts between them, see Data Types.

Inline references

Sometimes you don’t want to draw an edge, you want to drop a node’s output value into the middle of a prompt. Type @ inside any text field. A picker opens listing outputs you can reference, grouped into Connected to this node, Elsewhere on the canvas and Secrets. Keep typing to filter, then pick one and it drops into your text as a chip. When you run the node, the chip is replaced with the live value from the other node.
The @ reference picker open inside a prompt field
A reference chip inserted into a prompt after selection
Hover or click a chip to preview its current value. From there, Go to node selects and zooms to the node it points to, which is useful when the canvas is busy. Replace and Remove swap or delete the reference. For the full mechanics of references, see Data Referencing.

Showing parameters as inputs

Most node settings live in the node’s More (…) menu under All parameters. Click Show as input next to any parameter to turn it into an input row with a handle you can connect to from another node. Click Hide input to tuck it back into the menu.
The All parameters menu with Show as input next to a parameter
An Agent node with reasoning_effort shown as an input row with its own handle
This is how you make settings like aspect ratio, number of variations, or model temperature controllable from the outside, driven by:
  • A number input or slider
  • An agent’s output
  • Another node’s result
Showing a parameter as an input is separate from publishing. When you publish a workflow as a Tool or App, the Publish panel decides which inputs are exposed to the people and agents that use it. With the panel open, click the puzzle-piece icon on an input handle to add it to, or remove it from, the published inputs.

When to use which

  • Edges for primary data flow (the prompt, the image, the result).
  • @ references for values you want to inline into a prompt or template, they keep your canvas tidier.
  • Parameters shown as inputs for settings you want driven by another node, an agent, or an end user of your App.