EverArt MCP Server

✓ verified

AI image generation using multiple model backends.

Don't have ToolPiper? We'll take you to the download.

Open repository →

Publisher

Anthropic Reference (archived)

License

MIT

Transport

Local subprocess

Tools

1

What this server does

One tool, `generate_image`, with three parameters: a text `prompt`, a `model` ID, and an optional `image_count`. The server submits the prompt to EverArt's API, EverArt routes it to a worker for the chosen model, and the result URL comes back. The image opens in your default browser automatically. All outputs are 1024x1024. Five models are available without extra config: 5000 (FLUX1.1, standard photoreal), 9000 (FLUX1.1-ultra, higher fidelity), 6000 (SD3.5, Stable Diffusion 3.5), 7000 (Recraft-Real, photo style), and 8000 (Recraft-Vector, vector and illustration style). The default model ID `207910310772879360` is a specific EverArt-hosted checkpoint; pass one of the five IDs above to switch backends. Pricing is per-image against your EverArt account. The server does not gate or queue, so a model in a loop will rack up charges quickly. This is the simplest possible image-generation surface. Useful for quick mockups, blog thumbnails, or proof-of-concept generations inside a chat workflow. For control over resolution, aspect ratio, negative prompts, LoRA loading, or img2img, you want the model providers directly (Stability AI, FAL, Replicate) and a richer server. The upstream reference is archived and EverArt has not published a replacement, so this is the only MCP entry point to EverArt today.

Tools included

  • generate_image

    Generate an image from a text prompt using a selected EverArt model.

The authoritative list comes from the server's tools/list at runtime. This catalog entry may lag the upstream until the next ToolPiper release.

Install in ToolPiper

ToolPiper handles the JSON configuration. Click Install, fill in any required tokens, and the server is ready in your chat.

Don't have ToolPiper? We'll take you to the download.

Install in another client

Install in another MCP client (Claude Desktop, Cursor, Continue, Cline)

Add this to ~/Library/Application Support/Claude/claude_desktop_config.json.

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

Example prompts

  • "Generate a vector illustration of a fox in a forest using Recraft-Vector."

  • "Make four variants of 'minimalist iPhone wallpaper, pastel gradient' with FLUX1.1-ultra."

  • "Create a hero image for a blog post titled 'Local AI on Apple Silicon'. Use SD3.5."

  • "Generate a photorealistic portrait of a barista in a Tokyo cafe."

Security notes

The EverArt API key bills against your account on every successful generation. There is no spend cap inside the server; a model that loops on image generation will rack up charges fast. Keep the key in env, not in a config file checked into git. EverArt hosts the generated images at a public storage URL returned by the API; treat the URL as semi-secret if the prompt contained anything sensitive (faces, brand names, internal product mocks). The upstream Anthropic reference is archived and EverArt has not published an official MCP replacement. For new installations, consider providers with MCP servers that expose more controls, such as Stability AI, FAL, or Replicate.

Related tools