Image Generation MCP Server

@dfeirstein

About Image Generation MCP Server

Model Context Protocol (MCP) server enabling AI clients (Cline, Claude Desktop) to generate images using OpenAI (DALL-E 3, gpt-image-1) and save them directly to the user's project directory.

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "image-creator-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@dfeirstein/image-server@latest"
      ]
    }
  }
}

Tools

Generates an image using OpenAI (DALL-E 3 / gpt-image-1) based on a detailed text prompt. For best results, provide vivid descriptions incorporating style, composition, lighting, and mood. Can also edit or combine existing images by providing referenceImagePaths. Refer to 'docs/prompt-recipes.md' for extensive examples, templates, and tips for various image types (hero backgrounds, icons, illustrations, photos). Key parameters include 'prompt', 'brandSignature' (use project palette), 'size' (e.g., 1024x1024, 1536x1024), 'quality', 'model', 'filename', 'outputPath', 'targetProjectDir', and 'referenceImagePaths' (for editing/combining images).

Overview

What is Image Generation MCP Server?

An MCP server that allows compatible clients (like Cline, Claude Desktop) to generate images using OpenAI's API (gpt-image-1, dall-e-3, dall-e-2). It provides a create_image tool that takes a text prompt and other parameters, generates the image, and saves it to a specified project’s public directory.

How to use Image Generation MCP Server?

Install via NPX (npx -y @dfeirstein/image-server@latest) or clone the repository and run node dist/index.js. Configure your MCP client’s settings file (e.g., claude_desktop_config.json) with the server command and your OpenAI API key. The server exposes the create_image tool and serves documentation as MCP resources at docs/prompt-recipes and docs/readme.

Key features of Image Generation MCP Server

Generates images from text prompts using OpenAI models.
Edits and combines existing images via reference images.
Supports optional branding guidelines via brandSignature.
Allows specifying image size, quality, and background type.
Supports consistent visual styling through styleDefinitionJSON.
Saves images to a target project’s public directory.

Use cases of Image Generation MCP Server

Create logos, icons, or marketing assets with transparent backgrounds.
Generate themed image collections with a unified style definition.
Edit existing project images by applying effects or combining visuals.
Automate image creation inside an LLM-assisted development workflow.

FAQ from Image Generation MCP Server

What OpenAI models are supported?

The server supports gpt-image-1, dall-e-3, and dall-e-2. Image editing operations automatically use gpt-image-1.

What are the runtime requirements?

Node.js 18 or higher and npm are required. An OpenAI API key is mandatory; gpt-image-1 usage may need API Organization Verification.

How does the server save images?

Generated images are saved to the public folder of a target project directory specified by the LLM client. The server returns a relative path to the saved file.

Can I auto‑approve tools without confirmation?

Yes. In the client’s MCP settings, the autoApprove array can list tool names (e.g., ["create_image"]) to skip manual approval.

What transport and authentication does the server use?

The server uses stdio transport by default when launched by an MCP client. Authentication is via the OPENAI_API_KEY environment variable. A standalone HTTP server on port 5050 is available for testing.

Frequently asked questions

What OpenAI models are supported?

The server supports `gpt-image-1`, `dall-e-3`, and `dall-e-2`. Image editing operations automatically use `gpt-image-1`.

What are the runtime requirements?

Node.js 18 or higher and npm are required. An OpenAI API key is mandatory; `gpt-image-1` usage may need API Organization Verification.

How does the server save images?

Generated images are saved to the `public` folder of a target project directory specified by the LLM client. The server returns a relative path to the saved file.

Can I auto‑approve tools without confirmation?

Yes. In the client’s MCP settings, the `autoApprove` array can list tool names (e.g., `["create_image"]`) to skip manual approval.

What transport and authentication does the server use?

The server uses stdio transport by default when launched by an MCP client. Authentication is via the `OPENAI_API_KEY` environment variable. A standalone HTTP server on port 5050 is available for testing.

Comments

Basic information

More Media & Design MCP servers

GOJI

GOJI Agency

A remote MCP server from GOJI, a full-service digital agency in Melbourne, Australia. Nine tools over published material: a 104-term plain-English glossary, a 14-chapter guide to AI visibility, seven long-form guides inc

Instagram + TikTok Downloader & Transcriber

IYIA

Instagram + TikTok content tools inside your AI: download HD posts/reels/carousels, view Instagram stories anonymously, AI-transcribe videos (100+ languages, SRT/VTT/TXT). Requires a Repost PRO subscription.

Uwear.ai

Run complete AI fashion photoshoots from ChatGPT, Claude, or Codex. Turn flat-lay, packshot, and mannequin product images into catalog-ready on-model photos and video with saved art direction, automatic QA, and catalog-s

Vibe Design System

mondaycom

🎨 Vibe Design System - Official monday.com UI resources for application development in React.js

657

uxgen

A library of conversion components for e-commerce, built on A/B tested mechanics and handed straight to your coding agent. 65 pieces drawn from mechanics that have been A/B tested in real stores: cart drawer with the ups