MeiGen-AI-Design-MCP MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

1,500+ curated prompts with visual previews
Prompt enhancer converts rough ideas into production-grade prompts
ComfyUI workflow templates managed from the agent
Backend-agnostic: local ComfyUI, MeiGen Cloud, or OpenAI-compatible

Démo en direct

Aperçu en pratique

meigen-ai-design.replay ▶ prêt

0/0

Installer

Choisissez votre client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "meigen-ai-design": {
      "command": "npx",
      "args": [
        "-y",
        "MeiGen-AI-Design-MCP"
      ],
      "_inferred": true
    }
  }
}

Ouvrez Claude Desktop → Settings → Developer → Edit Config. Redémarrez après avoir enregistré.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "meigen-ai-design": {
      "command": "npx",
      "args": [
        "-y",
        "MeiGen-AI-Design-MCP"
      ],
      "_inferred": true
    }
  }
}

Cursor utilise le même schéma mcpServers que Claude Desktop. La config projet l'emporte sur la globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "meigen-ai-design": {
      "command": "npx",
      "args": [
        "-y",
        "MeiGen-AI-Design-MCP"
      ],
      "_inferred": true
    }
  }
}

Cliquez sur l'icône MCP Servers dans la barre latérale Cline, puis "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "meigen-ai-design": {
      "command": "npx",
      "args": [
        "-y",
        "MeiGen-AI-Design-MCP"
      ],
      "_inferred": true
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "meigen-ai-design",
      "command": "npx",
      "args": [
        "-y",
        "MeiGen-AI-Design-MCP"
      ]
    }
  ]
}

Continue utilise un tableau d'objets serveur plutôt qu'une map.

~/.config/zed/settings.json

{
  "context_servers": {
    "meigen-ai-design": {
      "command": {
        "path": "npx",
        "args": [
          "-y",
          "MeiGen-AI-Design-MCP"
        ]
      }
    }
  }
}

Ajoutez dans context_servers. Zed recharge à chaud à la sauvegarde.

claude mcp add meigen-ai-design -- npx -y MeiGen-AI-Design-MCP

Une seule ligne. Vérifiez avec claude mcp list. Supprimez avec claude mcp remove.

Cas d'usage

Usages concrets : MeiGen-AI-Design-MCP

Generate a batch of on-brand marketing images with MeiGen

👤 Marketers, solo founders, content creators ⏱ ~20 min intermediate

Quand l'utiliser : You need 10 variations of a hero image in a consistent style for an ad test.

Prérequis

A generation backend — Local ComfyUI on 127.0.0.1:8188, MeiGen Cloud token, or OpenAI-compatible image API key

Déroulement

Find a reference style

Search the MeiGen gallery for 'minimalist SaaS hero illustration'. Show me the top 5 with previews.✓ Copié

→ Gallery results with prompt IDs you can reuse
Enhance your specific idea in that style

Enhance this into a production prompt in the style of gallery entry #234: 'a workspace dashboard for a tax-filing tool, warm palette'.✓ Copié

→ Detailed prompt with style modifiers and negative prompts
Generate 10 variations

Generate 10 images from the enhanced prompt, varying the seed. Save to ./assets/hero/.✓ Copié

→ 10 PNGs in the folder

Résultat : A folder of on-brand variants ready for A/B testing.

Pièges

Cloud provider cost spike from large batches — Start batch at 2 to sanity-check quality, set a budget_cap via manage_preferences
Local ComfyUI out of VRAM on large batches — Reduce resolution or batch_size in the workflow template

Combiner avec : filesystem

Learn image prompting by iterating on the MeiGen gallery

👤 Designers new to AI image workflows ⏱ ~15 min beginner

Quand l'utiliser : You want to understand why certain prompts work before writing your own.

Déroulement

Browse the gallery by theme

Show me the top 10 trending prompts in the 'product photography' theme. Include the full prompt text.✓ Copié

→ Full prompts with structure visible
Dissect structure

For entry #5, break down the prompt into: subject, style, lighting, camera, modifiers. What makes it work?✓ Copié

→ Annotated breakdown

Résultat : You understand the anatomy of a working prompt — reusable knowledge.

Swap ComfyUI workflows from Claude without opening the UI

👤 ComfyUI power users ⏱ ~10 min advanced

Quand l'utiliser : You maintain multiple workflows (SDXL, Flux, inpaint) and want to switch based on task.

Prérequis

Local ComfyUI — github.com/comfyanonymous/ComfyUI; default port 8188

Déroulement

List saved templates

Use comfyui_workflow to list my saved workflow templates.✓ Copié

→ Templates by name and purpose
Load and run

Load the 'flux-portrait' template, inject my prompt, run and save the output.✓ Copié

→ Image saved; workflow JSON cached for reuse

Résultat : Faster iteration without leaving the chat.

Combinaisons

Associez-le à d'autres MCPs pour un effet X10

meigen-ai-design + filesystem

Generate and immediately organize into project folders

Generate 5 hero variants and save them under ./campaigns/{timestamp}/ with a manifest.json noting seed + prompt.✓ Copié

meigen-ai-design + claude-talk-to-figma

Design in Figma → refine with MeiGen → drop back into the Figma frame

Export the selected Figma frame's content brief, enhance with MeiGen, generate 3 options, import back into a new frame.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler	Coût
search_gallery	query: str, theme?: str, limit?: int	Find a starting style	0
get_inspiration	prompt_id: str	Expand a gallery hit into copyable prompt text	0
enhance_prompt	rough_idea: str, style_ref?: prompt_id	Convert a rough idea to a production prompt	0 or 1 API call
list_models		See what providers are configured	0
comfyui_workflow	action: list\|load\|run, name?, params?	Drive a local ComfyUI instance	Local compute
manage_preferences	style?, aspect?, model?, budget_cap?	Set defaults once	0
generate_image	prompt, negative?, aspect?, model?, count?	Actually produce images	Per-image, backend-dependent

Coût et limites

Coût d'exécution

Quota d'API: Gallery browsing: unlimited. Generation: depends on backend
Tokens par appel: Prompt ops: 500–3000. Image results: no tokens (paths only)
Monétaire: Free for browse/enhance; generation billed by chosen backend
Astuce: Iterate the prompt at low resolution (512px), only upscale the winner

Sécurité

Permissions, secrets, portée

Stockage des identifiants : Provider tokens in env vars (MEIGEN_TOKEN, OPENAI_API_KEY, etc.)

Sortie de données : Gallery: meigen.ai. Generation: configured provider endpoint only

Dépannage

Erreurs courantes et correctifs

generate_image returns 'no backend configured'

Run /meigen:setup or manually set one of: COMFYUI_URL, MEIGEN_TOKEN, OPENAI_API_KEY

Vérifier : list_models should show at least one provider

ComfyUI connection refused

Start ComfyUI (python main.py) and confirm it binds to 127.0.0.1:8188

Vérifier : curl http://127.0.0.1:8188/system_stats

Enhanced prompts are generic

Pass a style_ref from the gallery so the enhancer has an anchor

Alternatives

MeiGen-AI-Design-MCP vs autres

Alternative	Quand l'utiliser	Compromis
Direct ComfyUI MCP	You only want ComfyUI control, no gallery/enhancer	No curated prompts or cloud fallback
OpenAI Images API directly	You just want dall-e / gpt-image from code	No style library, no local option

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills