Web Design with an MCP/Agentic Model

I opened my Bluehost web hosting account today, and this popped up:

Screen capture of a Bluehost offer that asked to create a website fully with AI.

After chatting with Google’s Gemini about it, we came up with this:

The web ecosystem is rapidly shifting toward an agentic and Model Context Protocol (MCP) framework. However, how platforms like WordPress, Shopify, Wix, and Squarespace approach this transition depends heavily on whether they are closed “SaaS” systems or open-source ecosystems.

Instead of AI completely replacing these platforms, they are becoming the underlying “operating systems” that handle databases, checkout infrastructure, and hosting, while AI agents handle the operations, creation, and browsing.

A new way forward for designers:

  • Traditional web design means designers code lines, drag buttons, build pages piece by piece, and click through heavy dashboard menus.
  • The new “Agentic / MCP” model means designers ask, “Build me a store for my upscaled clothing with premium-priced branding”), and the AI handles all the mechanics under the hood (setting up databases, organizing products, and structuring the site).

In short: Instead of humans doing the tedious manual labor of building and managing websites, AI agents do the building and heavy lifting, while humans act as the creative directors telling the AI what to make.

The transition is unfolding across two distinct vectors:

1. Backend Operations (Where MCP is Winning Now)

The transition to MCP is already standard practice for store operations and management. Instead of users navigating complex dashboards, they use AI agents to manage their sites.

  1. Shopify and WooCommerce: Both platforms feature extensive MCP implementations. For example, Shopify offers native Storefront and Customer Account MCP servers, letting applications like ShopMCP enable merchants to use Claude or Cursor to update inventory, pull analytics, or edit themes in plain language.
  2. WordPress: While open-source WordPress has been slower to adopt a single native protocol, developers frequently use custom MCP setups to bypass the standard WP Admin completely, querying and updating database structures through external LLMs.

2. Frontend & Site Creation (The Rise of Agentic Builders)

The way websites are built and edited is shifting away from static drag-and-drop toward multi-agent collaboration.

  • Wix and Squarespace: Closed platforms are leaning into “vibe coding” and conversational design. Wix’s Aria agent (part of their Harmony ecosystem) acts as an autonomous co-creator rather than a rigid template generator. [1, 2]
  • WordPress Ecosystem: Agentic builders like 10Web deploy multiple specialized AI agents (e.g., a PM agent, a Designer agent, and a Developer agent) that collaborate to spin up production-ready WordPress sites in minutes. [1]

How Platforms Compare in the Agentic Shift

Shopify: The High-Tech Pro Store. It’s way ahead of the pack. It has built-in connections (“MCP”) so AI can easily manage inventory, process checkouts, and run the store automatically.

WordPress: The DIY Powerhouse. It doesn’t have one official AI system built-in, but because it’s open-source, you can plug in third-party AI tools (like 10Web) to automate development and site workflows.

Wix: The All-in-One Conversational Builder. Its built-in AI (called Aria) is baked right into the platform. You just chat with it to build the site and manage your dashboard.

Squarespace: The Good-Looking Designer. It uses basic AI mostly to help pick out stylish templates, layouts, and branding, but it’s not as deep into managing backend operations.

Why This Shift Is Inevitable

  1. The “Agentic Web” Evolution: Consumers are increasingly using AI agents (like Claude, ChatGPT, or Perplexity) to shop and browse for them. A website that does not expose an MCP server or structured API endpoint to external AI agents will eventually lose traffic because those agents won’t be able to read its product lines or content.
  2. Standardization: MCP provides a universal adapter. Instead of Shopify or WordPress writing custom integrations for every new LLM that hits the market, they provide an MCP server, and any AI agent can instantly understand the website’s data structure.

Impact on Graphic Design Education

The shift toward an agentic, prompt-driven web ecosystem fundamentally changes what graphic design and web design students need to master. For years, digital design curricula have focused heavily on manual software execution—pixel-pushing layouts in Figma, manually building CSS grids, and configuring rigid CMS dashboards template by template.

As AI agents and protocols like MCP take over the mechanics of site creation, hosting configuration, and layout assembly, education is shifting away from how to build the page manually toward how to art-direct the system intelligently.

Here is how this transition is reshaping design and web education across key areas:

1. From “Pixel-Pusher” to “Art Director of AI Systems”

When tools like 10Web, Wix Aria, or agentic coding assistants can spin up a fully responsive, structured multi-page website in minutes based on a conversational prompt, students who only know how to execute basic code or drag-and-drop elements will struggle to differentiate themselves.

  • The Shift: Students must transition from manual laborers to art directors and prompt architects. Success relies on knowing how to rigorously brief an AI agent on brand guidelines, color theory, typography hierarchy, and UX flow, and then critically evaluating and refining what the agent generates.

2. Teaching Systems and Constraints Over Single Pages

AI agents are brilliant at generating content quickly, but they often lack innate brand intuition or deep cultural context unless constrained properly.

  • The Shift: Curriculum design must double down on design systems, brand books, and rigid constraints. Students need to learn how to establish strict design parameters—such as defined token variables, precise tone-of-voice guidelines, and structured visual hierarchies—that they can feed into AI workflows to ensure the agent outputs cohesive, professional, and on-brand results rather than generic AI slop.

3. Elevating Strategic Thinking and Content Architecture

Because writing code, setting up databases, and importing products (like managing 44 digital art prints via CSV or API) can now be handled by conversational agents or automation protocols, the technical barrier to entry has collapsed.

  • The Shift: The value of a designer shifts entirely upstream toward strategy, user research, and narrative architecture. Students must spend more time mastering the foundational “why”—defining target personas, mapping user journeys, optimizing search taxonomy, and crafting compelling brand stories—because the AI will handle the “how.”

4. Preparing for API and Data-Driven Design

As storefronts and portfolios integrate with external AI agents and LLMs via MCP, the traditional boundaries between “graphic design” and “data structure” are dissolving.

  • The Shift: Designers need to understand how visual assets interact with backend systems. For instance, when designing layouts for digital downloads or e-commerce, students must consider how alt text, image metadata, structured schema, and API naming conventions affect how an external AI shopping agent perceives, categorizes, and recommends their client’s products.

Teaching HTML and CSS – Still Needed?

If your students are graduating into roles where they manage pre-built platforms (like WooCommerce, Shopify, or Wix) for local businesses, they rarely—if ever—write raw HTML or CSS from a blank text editor.

If your curriculum’s goal is strictly customization and styling tweaks within existing platforms, your current approach of teaching students how to inspect elements and inject custom CSS is entirely practical and aligned with the industry.

However, looking at where web design is moving over the next few years, the gap or what is shifting isn’t about writing code anymore—it’s about how we interact with the platform when things break or when we want custom features.

Here is what is changing, even in a pure CMS/SaaS environment:

1. From CSS Tweaks to Natural Language “Vibe Tweaks”

Traditionally, when a student wanted to change a button color or fix a mobile padding issue, they right-clicked, inspected the HTML class, and wrote a snippet of CSS into the customizer.

  • The Shift: AI assistants and platform agents (like Wix Aria or Shopify’s sidekicks) are taking over that workflow. Instead of hunting for .woocommerce-loop-product__title in the inspector, users will increasingly just highlight the element or type to the platform’s AI: “Make this product title bolder and move it above the image on mobile.” The AI writes the underlying CSS rules automatically.

2. When Plugins Don’t Just Work (The Integration Gap)

You mentioned that on WordPress, plugins either work or they don’t. That is true for standard, out-of-the-box features. But the moment a small business client wants something specific—like syncing their WooCommerce digital art inventory with an external AI tool, setting up a custom webhook, or connecting an automated email workflow—standard plugins often hit a wall.

  • The Shift: This is where the “Agentic / MCP” shift matters. Even low-tier users are starting to use lightweight automation tools (like Make.com, n8n, or free API keys) to make different software talk to each other. Understanding how data moves via APIs and JSON payloads is becoming the new “inspect element.”

3. The Changing Definition of “Design”

If platform tools and AI agents can handle layout adjustments, CSS tweaks, and template assembly instantly, the baseline value of a web designer shifts entirely.

  • What stays valuable: Brand consistency, asset curation (like your digital art mockups), understanding user behavior, and structuring clear content.
  • What becomes commoditized: Knowing how to find a div tag and change a hex code, because an AI can do that natively in seconds.

Summary: Do You Need to Change Much Right Now?

If your students are small business generalists who just need to launch a functional store, pick a good theme, tweak some styles, and manage content, you aren’t missing much at all. Your current approach is pragmatic.

Where you might want to future-proof them slightly is introducing them to AI-assisted styling and prompt-driven customization alongside traditional CSS inspection, so they aren’t left behind as platform dashboards turn into conversational chat boxes.