TL;DR: Tsukuru (作る) is a Model Context Protocol server that brings intelligent asset generation directly to your AI agents. Generate web images, iOS app icons, and marketing materials using DALL-E 3 with GPT-4 powered orchestration—all without leaving your development workflow.
Tsukuru means “to create” in Japanese, and that’s exactly what this tool does: it creates the visual assets you need, when you need them, powered by AI.
What Tsukuru Can Do
Tsukuru provides two powerful tools that transform how developers handle asset creation:
🖼️ Web Image Generation
Generate web graphics, banners, and marketing images using DALL-E 3 with intelligent optimization for different formats and use cases.
📱 iOS App Icon Sets
Create complete iOS app icon sets following Apple’s design guidelines, generating all required sizes and the proper Contents.json file.
Each tool is designed with developer workflows in mind. Need a hero banner for your landing page? Just describe it. Building an iOS app and need icons in all the required sizes? One prompt generates everything from App Store listings to Settings icons.
The magic happens through intelligent orchestration—describe what you need, and Tsukuru figures out which tool to use and how to configure it.
Design Philosophy
Intelligent Tool Selection
One of Tsukuru’s key innovations is rule-based orchestration. Instead of forcing developers to remember exact tool names and parameters, you can simply describe your needs:
- “Generate app icons for my iOS fitness tracker” → Automatically uses
ios-icontool - “Create a hero banner for my SaaS landing page” → Routes to
web-imagewith optimal settings - “I need marketing graphics for social media” → Configures appropriate dimensions and formats
Format Flexibility
Modern web development demands multiple image formats. Tsukuru automatically generates PNG, JPG, and WebP versions of every image, ensuring optimal performance across different use cases:
- ✅ High-quality originals for print and design
- ✅ Compressed WebP for fast web loading
- ✅ Standard formats for maximum compatibility
Apple Design Compliance
iOS app icon generation isn’t just about resizing—it’s about following Apple’s strict guidelines. Tsukuru generates complete icon sets with:
- ✅ Proper naming conventions
- ✅ All required sizes (from 20x20 to 1024x1024)
- ✅ The
Contents.jsonfile that Xcode expects
For more insights on building robust MCP tools, the MCP specification provides excellent guidance.
The Developer Experience
Seamless Integration
Tsukuru works with any MCP-compatible client. Whether you’re using Claude Desktop, Cursor, or building your own agent, setup is straightforward:
{
"mcpServers": {
"tsukuru": {
"command": "node",
"args": ["/path/to/tsukuru/dist/index.js"],
"env": {
"OPENAI_API_KEY": "your_openai_api_key_here"
}
}
}
}
Smart Defaults, Full Control
Tsukuru ships with sensible defaults but lets you customize everything:
- Image sizes: From social media squares to ultra-wide banners
- Quality levels: Standard for quick iteration, HD for final assets
- Output formats: PNG for transparency, JPG for photos, WebP for web
- File organization: Timestamped directories keep your assets organized
Example Workflows
Web Graphics Generation
# Through your MCP client
"Generate a modern hero image for a tech startup website featuring cloud computing and data analytics"
# Generates:
# temp/web-images-2024-06-09T15-30-45/
# ├── generated-image-1024x1024.png
# ├── generated-image-1024x1024.jpg
# └── generated-image-1024x1024.webp
iOS App Icon Sets
# iOS app icons
"Create iOS app icons for a meditation app with zen aesthetics"
# Generates complete icon set:
# temp/ios-icons-2024-06-09T15-30-45/
# ├── AppIcon-1024x1024@1x.png # App Store
# ├── AppIcon-60x60@3x.png # iPhone App
# ├── AppIcon-40x40@3x.png # iPhone Spotlight
# └── Contents.json # Xcode metadata
Technical Architecture
Tsukuru combines the best of modern development practices:
🔧 Modern Stack
- TypeScript: Full type safety with Zod validation
- MCP Compliant: Built with the official
@modelcontextprotocol/sdk - High Performance: Sharp-powered image processing
📊 Developer Experience
- Structured Logging: Pino-based logging for debugging
- Robust Error Handling: Graceful failures with helpful error messages
- Type Safety: End-to-end TypeScript for reliability
The architecture follows MCP best practices established by tools like Peekaboo MCP, ensuring reliability and compatibility across different AI agents.
My MCP Ecosystem Vision
Tsukuru is part of a growing movement toward specialized MCP servers that handle specific developer tasks:
- Peekaboo MCP - Screenshots and visual debugging
- claude-code-mcp - Task offloading to Claude Code
- Tsukuru MCP - Intelligent asset generation
- macos-automator-mcp - System automation
Each tool focuses on doing one thing exceptionally well, then composing together for powerful workflows.
The Vision: Assets as Code
A developer conducting AI agents like a symphony orchestra, visualizing the new role of system orchestration
Tsukuru represents a shift toward treating assets like code—version controlled, reproducible, and generated on demand. Instead of hunting through Figma files or stock photo sites, your assets live alongside your prompts in your repository.
The New Workflow
- Need to update your app’s branding? → Update the prompt and regenerate everything
- Launching in a new market? → Generate localized marketing materials
- Building a new feature? → Create the assets as part of your development process
When your design needs change, when your brand evolves, when you need assets for a new platform—instead of interrupting your flow to open design tools, you describe what you need and continue building.
That’s the power of bringing AI-powered asset generation directly into your development workflow.
🎨 Tsukuru MCP is available now - ⭐ star the repo if this speeds up your development workflow!
Built with ❤️ following MCP best practices. Tsukuru (作る) - because every developer deserves beautiful assets without the hassle.