Skip to main content
Run a complete workflow from code and save its output as first-result.png. This sample creates a 512 × 512 solid blue image using two built-in nodes. It verifies authentication, workflow execution, and output download without requiring a model, an input image, or custom nodes. To generate an image by calling a hosted model directly, start with the Comfy Router quickstart.

​
Before you start

This guide uses Comfy Cloud, the SDKs’ default target. If you previously set COMFY_BASE_URL, unset it to use Cloud. For an existing Comfy API deployment or a self-hosted instance, see Choosing a base URL.

​
1. Download the sample

Create a folder for the example. Download workflow_api.json and the script for your language into that folder: The workflow is already in API format. Node "1" creates the blue image and node "2" saves it. The scripts read outputs from that exact SaveImage node, so you do not need to find or edit a node ID.

​
2. Install the SDK and set your key

Open a terminal in the folder where you saved the files.
On Windows, activate the Python environment with .venv\Scripts\Activate.ps1 in PowerShell. Set your key in the same terminal:
Keep your API key in your server environment. Do not include it in browser code or commit it to source control.

​
3. Run and view the result

The script submits the workflow, waits for it to finish, and downloads the image. When it prints Saved followed by a path, open first-result.png from that location. You should see a solid blue square. Running the sample again replaces that local file.
This is a workflow connection check. It does not use a generative model. The same submit, wait, and download steps work with your own generation workflows.

​
If the request fails

  • Missing key or unauthorized: check that COMFY_API_KEY is set in the terminal running the script and contains an active key.
  • Access or credit error: check your Cloud subscription and available credits before retrying.
  • Workflow file not found: keep workflow_api.json next to the downloaded script.
  • Unexpected endpoint: check COMFY_BASE_URL. Unset it for this Cloud example.
For error handling and job progress, see the SDK guide.

​
4. Run your own generation workflow

Build or choose a workflow in the ComfyUI editor, run it successfully there, and export it in API format. Replace workflow_api.json with your export and update get_outputs("2") in Python or getOutputs("2") in TypeScript to use your workflow’s SaveImage node ID. Check that the target environment has the models and custom nodes your workflow uses. To host workflows with your own models and custom nodes, create a Comfy API deployment. To change workflow inputs, upload files, or watch progress, continue to Comfy SDKs.