ComfySharpSDK

ComfySharpSDK is a lightweight C# SDK for interacting with ComfyUI, a popular graphical interface for generating images using Stable Diffusion models. The library allows you to programmatically run workflows, track progress, and receive generated images directly in your .The NET application.

🚀 Quick Start

Did you like it? Put a star on it.

1. Launch ComfyUI

Make sure that your ComfyUI instance is running and accessible via HTTP (by default http://127.0.0.1:8188 ).
If you are using an API key (for example, when running with the --listen --enable-cors-header --extra-model-paths-config flag, etc.), specify it when creating the client.

2. Usage example

ComfySharpSDK/ComfySharpSDKExemple/RunWorkflowAndGetImagesAsync.cs

Lines 3 to 18 in d2b3bd4

	var client = new ComfyClient("http://192.168.0.10:8000");
	var workflow = Workflow.LoadFromFile(Environment.CurrentDirectory + "\\Workflows\\Default.json");
	workflow.SetNodeValue("6", "text", "Green Cat");
	var images = await client.RunWorkflowAndGetImagesAsync(workflow,
	(string message) => Console.WriteLine($"[PROGRESS] {message}")
	);
	foreach (var image in images)
	{
	if (image.Data == null || image.Filename == null) continue;
	if (!Directory.Exists("Outputs")) Directory.CreateDirectory("Outputs");
	await File.WriteAllBytesAsync(Path.Combine("Outputs",image.Filename), image.Data);
	Console.WriteLine($"Image saved as {image.Filename}");
	}

---

🧩 Main components

ComfyClient

The main client for interacting with ComfyUI.

Constructors

Basic constructor

ComfySharpSDK/ComfySharpSDK/ComfyClient.cs

Line 17 in 72bb66a

public ComfyClient(string baseUrl = "http://127.0.0.1:8000", string? apiKey = null, int maxAttempts = 120)

Constructor with custom HttpClient (for DI or setting timeouts)

ComfySharpSDK/ComfySharpSDK/ComfyClient.cs

Line 30 in 72bb66a

public ComfyClient(string baseUrl, HttpClient httpClient)

• baseUrl is the ComfyUI server address (without the ending /).
• apiKey is an optional Bearer token for authorization (if ComfyUI is protected).
• maxAttempts — the maximum number of attempts to wait for a result (120 by default, i.e. ~2 minutes with a delay of 1 sec).

Methods

Method	Description
RunWorkflowAsync(Workflow)	Sends workflow to ComfyUI and returns `prompt_id'.
GetHistoryAsync(string promptId)	Gets the execution history by `prompt_id'.
RunWorkflowAndGetImagesAsync(...)	Full cycle: start → wait → load images.

---

Workflow

Represents the ComfyUI workflow in JSON format.

Creation

// From a JSON string
var workflow = Workflow.Load(jsonString);

// From the file
var workflow = Workflow.LoadFromFile("path/to/workflow.json");

Setting up the parameters

Native way

workflow.SetNodeValue("6", "text", "a dog in space");

Important: node_id is a string identifier of the node in the workflow JSON structure (for example, "6").

The names of inputs (input_name) depend on the node type (for example, "text", "seed", "steps", etc.).

Fluent

You can freely change the node parameters by specifying the key and value for them

ComfySharpSDK/ComfySharpSDKExemple/FluentWorkflow.cs

Lines 9 to 22 in abadac3

	workflow
	.SetNodeValue("6", "text", "Green Cat")
	.WithNodeByTitle("Positive Prompt", inp => inp["text"] = "Green Cat")
	.WithNodeByClassType("CLIPTextEncode", inp => inp["text"] = "Green Cat");
	workflow.WithNodeByClassType("KSampler", inp =>
	{
	inp["seed"] = 156680208700286;
	inp["steps"] = 28;
	inp["cfg"] = 7.5f;
	inp["sampler_name"] = "euler";
	inp["scheduler"] = "normal";
	inp["denoise"] = 1;
	});

DSL

This is a more elegant way, but it also requires specifying the key and value types

ComfySharpSDK/ComfySharpSDKExemple/DslWorkflow.cs

Lines 9 to 20 in abadac3

	workflow.KSampler()
	.Set("steps", 30)
	.Set("seed", 12345)
	.Set(new {
	cfg = 7.5f,
	denoise = 0.9f
	});
	workflow.ForNode(6)
	.Set(new {
	text = "Cat"
	});

Dynamic Dsl

The most advanced way to edit a Workflow, the disadvantages are the need to download information about all nodes from ComfyUI, after which the system automatically collects objects from them

ComfySharpSDK/ComfySharpSDKExemple/DynamicDslWorkflow.cs

Lines 10 to 25 in abadac3

	dynamic sampler = await workflow.ForNodeAsync("KSampler", client);
	sampler
	.Steps(30)
	.Cfg(7.5)
	.SamplerName("euler")
	.Scheduler("Kerras");
	await workflow.SetNodeAsync("KSampler", client, b =>
	{
	b.Steps(30).Cfg(7.5).SamplerName("euler");
	});
	await workflow.SetNodeAsync(6, client, b =>
	{
	b.Text("Pen Cat");
	});

Additional information

• GetNodes() — returns the dictionary Node ID → name (if specified in _meta.title).
• toJSON() — serializes workflow back to readable JSON.

---

GeneratedImage

A class representing the generated image.

Property	Description
NodeId	ID of the node that generated the image
Filename	File name (for example, "ComfyUI_00001_.png")
Subfolder	Subfolder (if used)
Type	Output type (usually "output")
Data	Bytes of the image (byte[])

---

Authorization

If your ComfyUI is running with an API key (for example, via `--api-key your_secret_key'), pass it when creating the client.:

var client = new ComfyClient(
    baseUrl: "http://your-server.com:8188",
    apiKey: "your_secret_key"
);

The client will automatically add the header:

Authorization: Bearer your_secret_key

---

⏱️ Timeouts and cancellations

The RunWorkflowAndGetImagesAsync method supports the CancellationToken, which allows you to cancel the wait correctly.:

var cts = new CancellationTokenSource(TimeSpan.FromMinutes(5));
var images = await client.RunWorkflowAndGetImagesAsync(workflow, cancellationToken: cts.Token);

If the result does not appear in maxAttempts (by default, 120 attempts × 1 sec = 2 min), a TimeoutException will be thrown.

---

📁 Workflow structure

ComfyUI expects workflow in JSON format, where each node has a unique string ID and inputs (inputs). Example of a node:

{
  "6": {
    "_meta": { "title": "CLIP Text Encode" },
    "class_type": "CLIPTextEncode",
    "inputs": {
      "text": "a cat",
      "clip": ["4", 1]
    }
  }
}

You can change the "text" via:

workflow.SetNodeValue("6", "text", "a dog in space");

---

🛠 Tips

• Make sure that ComfyUI is running with the --listen flag if you are not using localhost.
• For CORS requests, --enable-cors-header may be required.
• Use Workflow.GetNodes() for debugging — it will show all available nodes and their names.
• Images are stored temporarily on the ComfyUI server. Do not delay the download for a long time.

---

📜 License

MIT License — use it freely in personal and commercial projects.

---

Note: ComfySharpSDK is not an official ComfyUI product. This is a community-driven SDK.