Nodes/ComfyUI-FlowChain/Workflow Output (FlowChain ⛓️)
ComfyUI Node

Workflow Output (FlowChain ⛓️)

The matching socket for a chained workflow's results

By numz·Created 2 years ago·Updated about a year ago· 170
Workflow Output (FlowChain ⛓️)
  • default
  • output
Name

WorkflowOutput is the back door of a FlowChain subworkflow - the mirror image of WorkflowInput. Where WorkflowInput declares what a workflow takes in, WorkflowOutput declares what it hands back. When you load the file as a subworkflow with the Workflow node, each WorkflowOutput in it becomes a named output socket on that node, so your chained workflow can return an image, a latent, a conditioning, or anything else you marked.

The whole FlowChain idea is "make a graph callable", and this node is what makes the call return something. If you've ever rebuilt the same ten-node pipeline by hand because you couldn't reuse it, this is the payoff: build once, mark the result with a WorkflowOutput, save, and the next graph treats the whole thing as a single box.

The inputs that matter

Only two, same shape as its sibling:

  • Name (STRING) - the output's label. It shows up on the Workflow node's socket, so name it by what it is (image, mask) rather than where it came from.
  • default (*) - the value that gets exposed. The author's description is short and accurate: "Default value is used to connect the output." You connect whatever you want the subworkflow to return into this input socket.

Type detection works in reverse of the input node: drag the WorkflowOutput's default input socket onto another node's output socket and the type auto-detects from what you touched. No type dropdown, nothing to mismatch.

The output

It also has an output socket named output (type *) and is flagged as an output node, so it both passes a value along and surfaces it in the UI. The order of multiple outputs is not arbitrary: the Workflow node sorts them by the vertical (top-to-bottom) position of the WorkflowOutput nodes inside the subworkflow, reading the real Y positions from the saved file so your socket order stays stable. Put them where you want them on the canvas and that's the order you'll see.

Installing it

Pack-level, nothing special. ComfyUI Manager → search FlowChain → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/numz/Comfyui-FlowChain.git
cd Comfyui-FlowChain
pip install -r requirements.txt

Then restart ComfyUI. No models, no weights - the pack's single pip dependency (gradio_client==0.8.0) only belongs to the LipSync node, so WorkflowOutput is dependency-free in practice.

Gotchas

A workflow won't appear in the Workflow node's dropdown until it has at least one WorkflowInput or WorkflowOutput - and since you typically need a marker on both ends, just remember the rule: mark the file before expecting to chain it. If you've changed which nodes are marked as outputs, the loaded node refreshes live (about a second after you save the subworkflow), but you may need to click back and forth between the two graphs to see it. And as with everything in this pack, keep ComfyUI at 0.3.33+ with a 1.18.9+ frontend; workflows saved by the old pre-2025 version of FlowChain won't convert.

CategoryFlowChain ⛓️

Inputs (2)

NameTypeDefaultDescription
NameSTRING
default*

Outputs (1)

NameTypeDescription
output*