N8n-nodes-comfyui-toolkit — non-blocking ComfyUI nodes with session tracking

Hey everyone! I just released n8n-nodes-comfyui-toolkit, a set of community nodes for integrating ComfyUI with n8n.

The problem with existing ComfyUI nodes

The available ComfyUI node for n8n blocks execution until generation completes. If you’re running 1000 concurrent requests, you end up with 1000 n8n executions all sitting in memory waiting — which doesn’t scale.

How this toolkit works differently

Generation is split into three separate steps:

  1. Submit (Text to Image / Image to Image / Text to Video / Image to Video / Upload Image) — sends the workflow to ComfyUI and returns prompt_id immediately. n8n is free.

  2. Wait Until Result — polls /history until the job completes. Run this only when you need the result.

  3. Get Results — downloads generated files and returns them as base64.

This means you can submit a job, hand off the prompt_id to your backend or a queue, and poll later — without tying up n8n memory.

Session ID system

Each generation node accepts a session_id which is passed through to all downstream outputs. This makes it easy to correlate a ComfyUI job with a chat session, user ID, or request ID — useful for building chat-based AI pipelines.

Works with any ComfyUI workflow

No custom wrappers. Export your workflow using ComfyUI’s Save (API Format), paste the JSON into the node, done. Tested with Flux but works with any model or pipeline.

[Trigger] → [Text to Image] → [Wait Until Result] → [Get Results]
[Trigger] → [Upload Image] → [Image to Image] → [Wait Until Result] → [Get Results]

Install

Settings → Community Nodes → n8n-nodes-comfyui-toolkit

Or:

docker exec -it n8n sh -c "cd /home/node/.n8n/nodes && npm install n8n-nodes-comfyui-toolkit"

Would love feedback — especially if you run into issues or have ideas for new nodes!