Block X · Content production at scale · Lesson 54
ComfyUI: from zero to a production workflow
You'll start with the desktop app and Comfy Cloud, get to know the paid nodes and learn to use Comfy MCP so an agent builds and runs the workflows for you.
- 11 min
- 9 animated slides
- 4-question quiz
- Narrated in 6 languages
What you'll see in this lesson
- ComfyUI: what it is and why it's used in production
ComfyUI is a free, open-source program where you build generation by connecting nodes, like Lego pieces.
- Install Comfy Desktop in ten minutes
The desktop app installs ComfyUI with its Python and dependencies: you download, run it and create your first install.
- Nodes, wires and workflows
A workflow is a chain of nodes connected by wires; each wire carries a data type and only fits where it belongs.
- Templates and models: start without building anything
The template library has ready workflows for MiniMax H3, LTX-2.5, Wan 2.2 or Qwen-Image and offers to download the models.
- Comfy Cloud: ComfyUI without your own GPU
Comfy Cloud is the same ComfyUI in the browser, with models and nodes preinstalled and NVIDIA RTX 6000 Pro GPUs.
- Paid nodes: Veo, Kling and MiniMax inside the workflow
Partner Nodes call closed models via API from your local ComfyUI and are paid with prepaid credits from your Comfy account.
- Comfy MCP: let the agent build and run workflows
Comfy MCP connects Claude Code, Claude Desktop, Codex or Cursor to ComfyUI: the agent finds templates, builds the workflow and runs it.
- Running via API: the move to production
Export the workflow in API format and send it to the ComfyUI server from a script, n8n or an agent, without opening the interface.
- Summary
ComfyUI brings all the models together in reproducible workflows you can run by hand, from an agent or via API.
Tools and concepts
- ComfyUI
- Comfy Cloud
- Comfy MCP
Sign in with your Google account to access this lesson and the whole course.