It's Monday. Here's what changed in AI this week, and what it means for your design work:

This week was all about onboarding a new client at the studio - Remy. He needs a whole new branding for his newsletter, and honestly, you should be reading it. He shares the AI tools and resources he actually uses in his own workflow, not the recycled lists everyone else posts. Follow it here -> aiwithremy.com

On the product side, we're about to ship a big update to Stage. You'll be able to add design skills and component libraries directly inside the tool, so what you generate comes out far more premium - less generic AI, more your taste. And because Stage is a native macOS app, it automatically detects the skills already installed on your machine. No setup, no copying files around. It just knows.

MCP just went from a nerdy side-experiment to the standard plug every creative tool speaks.

At SIGGRAPH last week (July 19-23), NVIDIA and nearly every major software maker made MCP - the Model Context Protocol - a standard feature across Adobe, Blender, Unreal Engine, Houdini, and Affinity. It's the first time MCP shipped as a real, built-in feature instead of a hacky proof-of-concept. Adobe went further and expanded its creative agent across Firefly, Express, and Creative Cloud, so you describe the outcome you want and the agent runs the whole multi-step job.

Never touched MCP? It's the layer that lets an AI agent like Claude actually control your software - open files, move layers, run steps - instead of just talking about it.

This is the quiet update that matters more than any new image model. MCP is becoming the USB-C of creative software - one standard connector every AI agent uses to plug into every tool. I've been running Claude + Figma this way for months, and it changes how you work: you stop clicking and start directing.

Once every app speaks the same protocol, your "workflow" stops being five disconnected tools. It's one agent moving between them while you handle taste and direction. The designers who win the next year won't be the ones with the flashiest tool - they'll be the ones who learn to direct agents across their whole stack.

Start small. Pick one tool you already use that has an MCP integration - Figma, Blender, or Photoshop via Adobe's agent - and connect Claude to it. Give it one boring, repetitive task (rename layers, build a grid, set up a scene) and watch it execute. That's your entry point.

You know the problem. You give an AI tool a prompt, and it hands back something that looks nothing like your brand - wrong colors, generic fonts, that unmistakable AI-slop feel. Then you burn 20 minutes re-prompting to get close.

We just rebuilt the Limora Brand Kit to kill that problem for good. Instead of describing your brand every time, you set it up once - and every asset you generate comes out on-brand automatically. It's now a full brand system, not a color picker. Paste your website URL and Limora re-scans it to pull your description, tagline, colors, and fonts as suggestions. Then you lock in your identity (logos, marks, industry, voice), your design language (primary/secondary/accent colors, heading and body fonts, tone tags like Minimal, Warm, Bold), and a Look & Feel section where you upload references and Limora extracts their style as editable tags - without ever storing your images. A final Conclusion section captures the essence of your brand, and what to never do.

You don't need to know Blender to make 3D anymore. With one free plugin, Claude writes the code and builds the scene for you - you just direct it.

💌 Know a designer who should be using AI smarter?
Forward them this email. Or just send them to logiaweb.net to join.

See you next Monday,

Say hi on X, Instagram or YouTube