Four nodes. Every Mage model. Your workflow, your way.
01
Install Mage in the Python environment ComfyUI uses, then restart ComfyUI. The nodes are in beta and their inputs may change.
Open ComfyUI-Manager, select Custom Nodes Manager, and search for Mage. Install it and restart when asked.
Open Comfy Registry02
Create an API key and set MAGE_API_KEY before starting ComfyUI. Or copy config.ini.example to config.ini and add your key there. It stays out of your workflows.
export MAGE_API_KEY="mage_sk_..."03
Find the Mage category and add a node to your graph. Each node shows its progress and Gem cost, and returns a URL alongside the result.
Generate or edit an image with Mango. Connect reference images, or the image to edit.
Generate a video with sound using Cherry. Connect reference images or a source video.
Voices, dialogue, music, or sound effects from one prompt, optionally matched to an image.
Every other Mage model. Set its fields as JSON and connect reference images, first and last frames, or a video.
Mention saved characters and references by @handle, just as you do on Mage. Save your favorite outputs within 30 days.
A fixed seed reuses ComfyUI’s cached result when your inputs stay the same. Change it whenever you want a new variation.
Every run uses your Mage Gems. Stopping the queue cancels the request; cancelled runs are not refunded.