The Death Warrant of Interfaces: MCP

The USB-C connector for AI tools: Model Context Protocol

That's a bold opening, but trust me, it's a modest beginning to explore what this protocol promises to deliver.

Now think about all the workflows you experience in daily life. How many different SaaS tools do you use? How many different operations do you perform? Even though we work in creative industries, a large portion of our daily operations are still operational.

At this stage, imagine that generative AI tools can connect their ability to produce things with the tools and services we already use. An LLM tool can access the tools you use with commands it receives from you and complete the operations you want to do. It does all of this through a single protocol. While it sounds a bit exciting, it can be difficult to imagine without seeing examples first.

Real-World Use Cases

Solution to Revision Fatigue

You're a designer and you've been able to cope with your clients' constantly changing demands until now. The revisions never ended. However, the scope of the changes made so far and the direction the platform is going are already clear.

Now imagine your LLM tool interpreting the emails you receive, compiling discussions in Slack channels, and listing clients' change requests in the project management tool you use.

So far so good, let's go further, imagine implementing the requested changes on Webflow by chatting with the LLM tool. While all this is happening, documenting the changes made in the project management tool is also taken care of. This is possible today, not tomorrow.

Content Creation and Distribution

Personally, the biggest question mark I had before I started producing content was that the process of producing and sharing content was taking too much time among ongoing work.

You researched or were exposed to something that excited you. You found it worth sharing.

Improving the content, moving it to a blog format.

Creating short clips from this is a long job.

Unfortunately, some of the tools I use still require manual operations, but it feels very close to being fully automated. I currently operate the following workflow:

During different times of the day, I save content I find interesting to a Knowledge Base table I created in Notion. From time to time, I visit here and write my own thoughts about the content in this table.

When I start the content production process, I discuss current topics I've collected in the Knowledge Base with Claude. I ask it to turn this into a blog post. Then I say publish this on my website.

Or I say create a video in Captions for the content I last published on the website.

I wish I could convert these into a Carousel via Figma. However, Figma's MCP capabilities are currently limited to converting designs to code or interpreting these designs. But it won't be long.

Useful MCP Solutions for Your Work

From Design to Code with Figma MCP

You can very quickly convert your designs to code by connecting Claude or Cursor to your Figma. You can watch this video by Ramazan Güler from Brick Institute trainers on this topic.

Designers are probably wondering "can it edit designs?" right now. No, folks. Unfortunately, it doesn't work that way. You can currently quickly convert your designs to code in Figma. However, you cannot create new designs from your existing components yet.

Tip
For this, you can produce outputs you can use directly in Figma with products like UX Pilot or Google's Stitch.

Expanding Boundaries with n8n, Zapier & Make

MCP allows you to integrate with tools that already have MCP Server setup with your AI tool, but to integrate with tools that don't have this capability yet, you can connect to n8n, Zapier and Make via MCP and set up automations with APIs to all other tools in these tools' integration networks. When you get here, there are actually no limits.

Industry Reflections and Interesting Examples

✈️ Turkish Airlines MCP Server: Even when LLMs weren't around yet, the THY team and Mindbehind were working together on the user experience of virtual assistant Boti. This was a good vision for that period too. Today, technology promises much more than that. This article explains how Turkish Airlines developed an MCP server that brings AI assistants together with real-time airline data.

How to Get Started?

I prefer Claude as the LLM tool here. I've prepared content on how to get started with this.

Example Process Flow Through This Article

In the video below, I show how I publish this article content on my website just by chatting with Claude.

Is the End of Interfaces Coming?

As designers and product managers, no matter how much we try to reduce, shorten, and simplify, the interfaces we create for using each tool have a learning curve and cognitive load. Let's admit that it's not very easy for anyone to create a good and easy-to-use product.

LLM tools have already changed our information acquisition process. A 30-40% decrease in internet traffic is foreseen. But this is just the beginning. It will also reduce our habit of using products. We will have an Assistant LLM + MCP connections working for us doing our job. It's turning into a world where LLMs are positioned between interfaces and users.

Maybe it's early to say that the end has come. But when you imagine the last year, it's likely that the screens you've seen the most are the ones above. With this interface structure, we can now edit videos, update websites, and convert our Figma designs to code.

For a more technical examination of how MCP works, you can watch this video by the "fatih loves to simplify" channel, which is a more technical review than this article.