This guide shows you how to create, test, and save your own ComfyUI workflows in RunComfy’s cloud environment, from building or modifying workflows, to exporting them for API use, to cloud-saving everything for deployment.

​
Overview

In RunComfy, you build workflows as node-based graphs in a cloud-hosted ComfyUI environment. When your workflow is ready, you can Cloud Save it, packaging the workflow and its full runtime (drivers, libraries, custom nodes, models, and dependencies) into a container image that runs consistently in the cloud every time. RunComfy’s ComfyUI Cloud is kept in sync with the official ComfyUI GitHub repository, including historical versions, so you get the same interface and tools you use locally, no relearning required. You can easily download models from Civitai, Hugging Face, or Google Drive, upload your own, and install custom nodes using the built-in ComfyUI Manager. Once saved to the cloud, your workflow can be deployed as a serverless API endpoint, scalable, production-ready, and covered in detail in the Create a Deployment section.

​
Build a Workflow

You can start with a clean environment or use an existing community workflow as your base.

​
From Scratch

To begin a ComfyUI session, head to the My Workflows page. You’ll find two preset environments to pick from and launch a machine. The ComfyUI-NodesLoaded option comes with popular nodes pre-installed to cut down on setup time. The ComfyUI-Minimal option skips preloads for quicker startup. Both draw from the latest ComfyUI version on GitHub, and you can choose older versions via the three-dot dropdown before launching. When the interface loads, either build your workflow from scratch or import a workflow JSON by dragging it onto the canvas. Alt Build a Workflow from Scratch

​
Add Custom Nodes

To install custom nodes, click the Manager button in the interface. Select Install Missing Custom Nodes to auto-detect and add what’s needed, or pick Install Custom Nodes and search for a specific node. After installation, restart ComfyUI using the Manager’s restart feature and refresh your browser to activate the changes. Alt Install Custom Nodes in ComfyUI

​
Add Models

Click the “Assets” button on the right sidebar to open the file browser panel. For downloading models from Civitai, Hugging Face, or Google Drive, first navigate to the appropriate subfolder (e.g., models/checkpoints or models/loras) in the browser. Enter the model’s URL in the download bar at the top, click Download, then refresh the interface to load the new model, it will save directly to the selected folder. Read more here: How to download models from Civitai, Hugging Face, and Google Drive? Alt Download Models in ComfyUI For uploading models from local storage, navigate to the target subfolder in the file browser, click the three-dot menu, select Upload, and choose your files, they will upload directly to the selected folder Read more here: How to Upload/Delete/Move files in RunComfy? Alt Upload Models in ComfyUI

​
Use a Template

To pick a community workflow, visit the Explore page, browse or search for one that fits, and click Run Workflow to load it into a machine instance. When the ComfyUI interface loads with the pre-built workflow, make changes as required, using the same steps above for installing nodes or models. Alt Use a Workflow Template

​
Test Workflow

With your workflow set up, click Queue Prompt to run it and preview outputs (images, videos, etc.) directly in output nodes. Files are automatically saved to the output folder. For debugging during tests, you can search for information in the node author’s related GitHub repositories, or refer to the common error fixes in RunComfy’s How-tos.

​
Export for API

When your workflow is stable, export it as API JSON by selecting Export (API) from the top-left Workflow menu. This generates a JSON file containing all nodes, inputs, default values, and connections, which you can download or copy for later use. Alt Export API in ComfyUI The exported API JSON is structured like a standard ComfyUI workflow JSON: it’s a dictionary where keys are node IDs (e.g., "6" for a prompt encoder or "25" for a sampler), and each value is a node object with its class_type (the node name), inputs (holding default values or links to upstream nodes), and _meta for extras like titles. Links in inputs appear as arrays like ["11", 0], meaning “connect to output slot 0 from node 11”. Here’s a snippet of two nodes from a sample FLUX workflow: 
{
  "6": {
    "inputs": {
      "text": "An old tv with the word \"FLUX\" on it, sitting in an abandoned workshop environment, created in Unreal Engine 5 with Octane render in the style of ArtStation.",
      "clip": ["11", 0]
    },
    "class_type": "CLIPTextEncode",
    "_meta": {
      "title": "CLIP Text Encode (Prompt)"
    }
  },
  "25": {
    "inputs": {
      "noise_seed": 417264191485848
    },
    "class_type": "RandomNoise",
    "_meta": {
      "title": "RandomNoise"
    }
  }
}
View the complete example here: runcomfy-flux-workflow-api.json.

​
Apply Overrides

Overrides allow you to tweak specific parts of a deployed workflow right in your API calls without having to send the full workflow JSON every time. Just add them to the body of your POST request to the /inference endpoint, which lets you make on-the-fly changes like swapping prompts or seeds based on user input, keeping requests fast and your app scalable. To build overrides, create a JSON object where keys are node IDs from your exported API JSON, and each value is an object with an inputs key containing the new values for that node’s parameters. Only override what needs changing; the workflow’s defaults apply elsewhere. For media like images, use public URLs or Base64 data URIs in the input fields. Here’s an example updating the prompt text and noise seed from the sample workflow: 
{
  "overrides": {
    "6": {
      "inputs": {
        "text": "futuristic cityscape"
      }
    },
    "25": {
      "inputs": {
        "noise_seed": 987654321
      }
    }
  }
}

​
Cloud Save Workflow

In RunComfy, Cloud Saving packages your entire ComfyUI workflow, along with its runtime environment, drivers, libraries, custom nodes, models, and dependencies, into a fully reproducible container image. This ensures your workflow runs consistently in the cloud, regardless of the underlying hardware or environment. Cloud Saving keeps workflows deployment-ready, supports versioning for iterative updates, and enables private sharing within your team, so you can collaborate smoothly without worrying about dependency conflicts.
Note: Community workflows in RunComfy, found in the Explore page, are already pre-saved with Cloud Saving, so you can use them immediately or modify and save them as your own.

​
Save to Cloud

Save the workflow by clicking the Cloud Save button in the top bar. This packages the entire setup, including all dependencies into a container image that’s ready for consistent cloud execution.
Note: If multiple workflow tabs are open, only the active tab will be saved. Switch to other tabs and save them separately as new workflows to preserve each one.
Alt Save Workflow to Cloud in RunComfy Read more here: RunComfy ComfyUI Workflow Cloud Save and Sharing features

​
View Saved

Saved workflows appear on your My Workflows page. Next time you run it, you’ll get the saved version complete with its environment. Alt Saved Workflows

​
Deploy Saved

Saved workflows are ready to deploy right away: click the API button or go to the Deployments page, select your workflow, and set up the endpoint. 🎉 Your workflow is now built, tested, and cloud-saved, ready to become a production API!