ComfyUI MCP Server
@zuojianghua
ComfyUI MCP Server について
generate_image and other workflows
基本情報
設定
以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。
{
"mcpServers": {
"ComfyUI_MCP_Server": {
"disabled": false,
"timeout": 600,
"command": "python",
"args": [
"D:\\code\\comfyui_mcp_server\\server.py"
],
"env": {
"COMFY_URL": "http://127.0.0.1:8188/"
},
"transportType": "stdio"
}
}
}ツール
ツールは検出されませんでした
ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。
概要
What is ComfyUI MCP Server?
A Model Context Protocol (MCP) server for ComfyUI that provides image generation and prompt optimization services.
How to use ComfyUI MCP Server?
Configure the server in your MCP client using the provided JSON snippet, setting the COMFY_URL environment variable to your running ComfyUI instance (default http://127.0.0.1:8188/). The server runs as a Python script via stdio transport. Ensure Python 3.7+ and the FastMCP library are installed, and that a ComfyUI instance is active.
Key features of ComfyUI MCP Server
- Image generation from text prompts using ComfyUI workflows
- Prompt optimization to improve generation results
- Automatic image dimension adjustment (multiples of 8)
- Random seed generation for diverse outputs
- Returns both local file paths and online accessible URLs
Use cases of ComfyUI MCP Server
- Generate images from textual descriptions via ComfyUI’s text_to_image workflows
- Enhance prompt quality to achieve better image generation results
FAQ from ComfyUI MCP Server
What are the requirements to run the server?
Python 3.7+, a running ComfyUI instance, and the FastMCP library. The ComfyUI API must be accessible at the URL specified in COMFY_URL.
How does the server handle image dimensions?
It automatically adjusts image dimensions to multiples of 8 to ensure compatibility with ComfyUI’s model requirements.
What transport does the server use?
The server uses the stdio transport, as shown in the default MCP configuration example.
Are results returned as files or URLs?
Both local file paths and publicly accessible URLs are provided for each generated image.
How are diverse outputs generated?
A random seed is generated per request, producing varied results from the same prompt.
「その他」の他のコンテンツ
Blender
ahujasidOpen-source MCP to use Blender with any LLM
Awesome Mcp Servers
punkpeyeA collection of MCP servers.
Core Philosophy: Connect, Unify, Respond
mindsdbDelegate anything. It comes back done.

Sequential Thinking
modelcontextprotocolModel Context Protocol Servers
ICSS
chokcoco不止于 CSS
コメント