AI导哈

ComfyUI图像生成服务器

一个基于ComfyUI API的图像生成服务器，支持多种图像处理功能，适用于自动化图像生成和处理的场景。

  • by LumiVerseHR
  • TypeScript
  • 本地部署
  • 更新于
在 GitHub 查看 LumiVerseHR/mcp-server-comfy-api

使用说明

@lumiversehr/mcp-server-comfy-api

MCP server for generating images with ComfyUI API using z-image-turbo, directly from Claude Code.

Quick Setup

Add to your Claude Code config (~/.claude.json under mcpServers):

"comfy-api": {
  "type": "stdio",
  "command": "npx",
  "args": ["-y", "@lumiversehr/mcp-server-comfy-api"],
  "env": {
    "COMFY_API_URL": "https://your-comfy-api-host.com",
    "COMFY_API_KEY": "your-api-key-here"
  }
}

Restart Claude Code. Done.

Authentication

The backend requires an API key for all requests. Set COMFY_API_KEY in the MCP server's env config to match the API_KEY value configured on the backend's .env file.

Requests without a valid X-API-Key header will be rejected with 401.

Tools

Tool Description
generate_image Generate a z-image from a text prompt. Waits for completion (up to 3 min). All images go to a private "API_generated" series.
get_job_status Check status of a generation job by ID.
download_image Download a generated image to the local filesystem.
remove_background Remove the background from a local image, saving a transparent RGBA PNG.
list_pipelines List saved node-graph pipelines that can be run.
get_pipeline_params List a pipeline's parameters (name, type, default).
run_pipeline Run a saved pipeline with named inputs (text/number/image); returns output image filename(s)/text. Use download_image to save them.
get_prompt_guide Retrieve the Z-Image Turbo prompt guide for writing effective prompts.

Usage Examples

From Claude Code:

  • "Generate an image of a sunset over mountains"
  • "Show me the prompt guide for z-image"
  • "Download that image to ./sunset.jpg"
  • "What's the status of job abc-123?"
  • "Remove the background from ./photo.jpg"
  • "List the comfy-api pipelines" → "Run IsometricTiles with tile = bathroom and download it"

remove_background Parameters

Parameter Type Default Description
image_path string required Path to the local input image (PNG/JPG/WebP)
output_path string <input>_nobg.png Where to save the transparent PNG
model enum u2net u2net (general), u2netp (lighter/faster), u2net_human_seg (people), isnet-general-use (higher quality), isnet-anime, silueta (portraits)
alpha_matting boolean false Cleaner edges (slower)

generate_image Parameters

Parameter Type Default Description
prompt string required Text description of the image
width number 1024 Image width (256-2048, rounded to 16x)
height number 1024 Image height (256-2048, rounded to 16x)
steps number 9 Diffusion steps (1-50, 9 recommended for turbo)
cfg number 1.0 CFG scale (turbo doesn't use CFG, keep at 1.0)
seed number random Seed for reproducibility

Configuration

Environment Variable Description Default
COMFY_API_URL Backend API base URL https://comfy-api.nichetide.com
COMFY_API_KEY API key for authentication (required)

How It Works

  1. generate_image submits a job to the backend API, polls every 2s until complete
  2. All generated images are auto-assigned to a private "API_generated" series
  3. Jobs run at priority 5 (higher than default web UI jobs)
  4. download_image fetches the image bytes and saves to your local filesystem
  5. remove_background uploads a local image, runs background removal, and saves the transparent PNG locally
  6. All requests include the X-API-Key header for authentication

Requirements

  • Node.js 18+
  • Network access to the ComfyUI API backend
  • Valid API key

License

MIT

相关 MCP