EverArt MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

Single tool, zero setup beyond an API key
Multiple models: FLUX1.1, SD3.5, Recraft variants
Returns a hosted URL — paste into Notion, commit to repo, or download
Note: repo is archived — publisher status unchanged but no new features

Démo en direct

Aperçu en pratique

everart.replay ▶ prêt

0/0

Installer

Choisissez votre client

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

{
  "mcpServers": {
    "everart": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-everart"
      ]
    }
  }
}

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

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "everart": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-everart"
      ]
    }
  }
}

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

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "everart": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-everart"
      ]
    }
  }
}

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

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "everart": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-everart"
      ]
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "everart",
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-everart"
      ]
    }
  ]
}

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

~/.config/zed/settings.json

{
  "context_servers": {
    "everart": {
      "command": {
        "path": "npx",
        "args": [
          "-y",
          "@modelcontextprotocol/server-everart"
        ]
      }
    }
  }
}

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

claude mcp add everart -- npx -y @modelcontextprotocol/server-everart

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

Cas d'usage

Usages concrets : EverArt

Generate OG / social card images for blog posts

👤 Content writers, solo devs running a blog ⏱ ~10 min beginner

Quand l'utiliser : You just wrote a post and need a 1200×630 social card before publishing.

Prérequis

EverArt API key — everart.ai → account → API keys. Set as EVERART_API_KEY.

Déroulement

Draft the prompt from the article

My post is titled 'How we cut p99 latency in half'. Propose 3 image prompts that capture that visually — abstract, not literal, tech-editorial style.✓ Copié

→ 3 distinct prompt candidates
Generate variations

Use EverArt to generate image_count=3 for the prompt I pick. Use the FLUX1.1 model for cleaner composition.✓ Copié

→ 3 URLs returned
Pick and save

I like #2. Download it to ./public/og/<slug>.png via the filesystem MCP.✓ Copié

→ File saved locally, ready for deploy

Résultat : A publish-ready OG card in 5 minutes, no design app open.

Pièges

Generated images have artifacts (weird text, extra fingers) — Skip prompts that ask for text in the image — let the model do abstraction and overlay your title via CSS/HTML
Aspect ratio doesn't match 1200×630 — The model outputs 1024×1024 by default — crop/resize via an image tool or accept square

Combiner avec : filesystem

Generate a landing page hero image

👤 Founders, marketers on tight budgets ⏱ ~15 min beginner

Quand l'utiliser : Pre-launch landing page, no designer, stock photos feel generic.

Déroulement

Describe the vibe

My product is <description>. Target audience: <persona>. Propose 3 hero image directions that would feel authentic.✓ Copié

→ 3 direction briefs
Generate 4 variants of direction #1

Generate 4 images for: '<chosen prompt>'. Use Recraft for clean illustration style.✓ Copié

→ 4 URLs
Iterate

#2 is closest. Tweak: warmer palette, less cluttered background. Regenerate 3 more.✓ Copié

→ Refined set

Résultat : A distinctive hero image for roughly the cost of a coffee.

Pièges

Over-iterating burns credits fast — Cap yourself at 2 iteration rounds; if nothing clicks, the direction is wrong — reset the brief

Combinaisons

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

everart + filesystem

Generate then save locally for commit

Generate an OG card for my 'caching' post, download the URL to ./public/og/caching.png.✓ Copié

everart + notion

Generate an illustration and embed it directly in a Notion page

Generate a banner image for the 'Onboarding' page and append it as an image block at the top.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler	Coût
generate_image	prompt: str, model?: str (default '207910310772879360'), image_count?: int (default 1)	The only tool — generate N images matching a prompt	1 EverArt credit per image

Coût et limites

Coût d'exécution

Quota d'API: Depends on your EverArt plan — usually measured in credits
Tokens par appel: Prompt text only (~100 tokens); generation time is the bottleneck
Monétaire: Check everart.ai/pricing — typically cents per image
Astuce: Generate image_count: 4 in one call instead of 4 separate calls — same cost, faster iteration

Sécurité

Permissions, secrets, portée

Stockage des identifiants : EVERART_API_KEY in env var

Sortie de données : Prompts and images to everart.ai

Repo is in modelcontextprotocol/servers-archived — functional but no active maintenance. Consider fal.ai, Replicate MCPs, or an official vendor MCP for production.

Dépannage

Erreurs courantes et correctifs

401 Unauthorized

EVERART_API_KEY missing or invalid. Re-create at everart.ai and set in your MCP client config.

Vérifier : curl -H "Authorization: Bearer $EVERART_API_KEY" https://api.everart.ai/v1/models

Image URL expires after a day

Download immediately with the filesystem MCP — don't rely on the hosted URL long-term.

Generation timeouts

Large image_count values can exceed client timeouts. Request 2 at a time and loop.

Alternatives

EverArt vs autres

Alternative	Quand l'utiliser	Compromis
Replicate MCP	You want broad model choice (FLUX, SDXL, hundreds of fine-tunes)	More setup, more powerful, pay-per-second pricing
fal.ai MCP	You need fast inference and streaming	Similar pricing; slightly fewer models than Replicate
OpenAI DALL-E via API	Already on OpenAI and want one bill	Stricter content filters; narrower style range

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills