Lexsis AI
Lexsis AI

Table of Contents

AI Storefronts

How to automate A/B testing with Shopify + Lexsis AI MCP + Storefront Skills and Claude Fable 5

10 min read
32 views

We open-sourced the skill packs we use internally to generate Shopify landing pages with Claude Code, OpenAI Codex, and Cursor.

One prompt in. Working page out. Add-to-cart, product data from your store, reviews section, sticky CTA, responsive mobile layout. Published to your own domain in about 60 seconds.

Works with Fable 5, Opus, Sonnet, Codex, GPT-5 - any model that supports tool use.

Repo:

https://github.com/Aetheria-Labs1/storefront-skill-packs

This is what a generated page looks like for a haircare brand:

Article content

Claude Fable with Lexsis AI MCP , Brand - Moxie Beauty India

What Skill Packs Actually Are

Skill packs are markdown files you drop into your AI coding agent's config directory. They contain structured knowledge about how to build high-converting Shopify pages - extracted from patterns across 500+ stores we've analyzed.

They cover four areas:

Conversion psychology - AIDA section ordering, price anchoring (strikethrough + current price), trust escalation ladders, CTA copy patterns, cognitive load limits (max 3 choices per section).

Vertical expertise - Beauty brands need ingredient storytelling and before/after layouts. Supplement brands need clinical proof sections and dark backgrounds. Fashion needs editorial grids. Each vertical has different patterns that convert.

Traffic source optimization - A page receiving Meta ad traffic needs message-match with the ad creative, mobile-first layout, and trust stacking near the fold. Google traffic expects information density and comparison tables. TikTok traffic needs a 3-second visual hook and UGC aesthetic.

Workflow orchestration - Which MCP tools to call in what order, what can run in parallel, when to check brand assets before generating new ones.

Available Packs

Article content

Skills Packs

If you prefer a video tutorial, watch it here:

Full Setup Guide (Zero to First Generated Page)

Step 1: Install Lexsis AI on Shopify

Find Lexsis AI in the Shopify App Store*. Click "Add app" and authorize.

Article content

Lexsis AI app on Shopify App Store

We are facing some issue with our app review at the moment , We will edit the link to the app here once its live , For now , you can DM us or Book a demo on our website and we can install this as a custom app on your store

After installing , You need to head to the Lexsis platform to create your account.

Step 2 : Create an account

You can create an account here -

https://app.trylexsis.com/sign-up?utm_source=twitter&utm_medium=social&utm_campaign=storefront_skill_packs_launch&utm_content=article

Step 3 : Select your workspace type

Article content

Choose between brand and agency mode.

After selecting ,fill up your actual brand details, you will see an option to connect to your Shopify domain.

Step 4: Connect Shopify

Link your store. This syncs your product catalog (images, variants, prices), collections, theme settings, and navigation menus into Lexsis.

If you face any issues in connection, please do reach out to us at

support@trylexsis.com or you can DM us here on X @trylexsis or @addddiiie

Once the connection succeds , check back on your shopify store for the catalogue sync and brand kit sync to trigger in the background (may take anywhere between 2 to 5 minutes)

Article content

The Lexsis app dashboard in your Shopify admin

Step 5: Brand Kit (Automatic)

Lexsis auto-generates a Brand Kit from your connected store:

  • Logo detection

  • Color palette extraction from your theme

  • Typography identification

  • Voice/tone analysis from existing product copy

This feeds into every page the AI generates. Your pages come out on-brand without manual theming.

Once your catalogue is synced you will find all your products visible in the dashboard

Article content

Store catalogue synced to Lexsis app

Article content

Design.md of your brand used as context

Article content

Design tokens extracted for theming

Step 6: Create an API Key

Go to Settings > API Keys. Click "Create API Key". Select admin permissions.

Copy this key. You'll paste it into your MCP config in the next step.

Step 7: Install the Skill Packs

 git clone https://github.com/Aetheria-Labs1/storefront-skill-packs
cd storefront-skill-packs/starter
./install.sh --target /path/to/your/project 

The installer detects whether you're using Claude Code, Codex, or Cursor and copies the appropriate files.

Step 8: Connect the Lexsis AI MCP Server

The MCP server gives your AI agent access to tools: image generation, page validation, publishing, brand kit fetching, product listing, ad creative analysis.

Claude Code

Create .mcp.json in your project root:

{
  "mcpServers": {
    "lexsis-ai": {
      "type": "http",
      "url": "https://mcp.trylexsis.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

For more supported platfroms , please check installation guide on

https://docs.trylexsis.com/docs/getting-started/authentication

Step 9: Generate Your First Page

This is my session using it with OpenAI Codex

Once you have installed the skills , you will verify to see if they are visible using / (slash) commandsOnce you have installed the skills , you will verify to see if they are visible using / (slash) commands

Article content

Codex view using / commands

You can also install industry specific skills according to your niche and are also welcome to contribute more

Article content

To test if your authenticated correctly with the MCP is just to ask it get your details

Prompt -

Connect to my Lexsis workspace using the Storefront MCP and enumerate all available tools, resources, prompts, and capabilities exposed by the server. Briefly describe what each tool does and identify any available storefront, Shopify, page management, or AI-related operations.

Find example prompts here - https://docs.trylexsis.com/docs/getting-started/quickstart

Once you have confirmed authentication you can ask it to create any page for the following use cases:

  1. From an ad creative: Upload your Meta ad image. The agent extracts the visual style, headline, claims, and color palette, then builds a landing page that continues that visual story post-click.

  2. From a competitor URL: Point the agent at a competitor's product page. It deconstructs the layout, identifies the conversion patterns, and rebuilds the page for your brand with your products and brand kit.

  3. From a brief: Describe what you sell, who buys it, and where the traffic comes from. The agent picks the right vertical patterns and traffic-source optimizations automatically.

Why Fable 5 Is Relevant Here

Fable 5 is Anthropic's fastest model. For this workflow, speed matters because:

  1. Skill packs front-load the design decisions. The model doesn't need to reason about what makes a good landing page - that knowledge is pre-loaded. It spends tokens on execution, not deliberation.

  2. The architecture is vibe-code. Raw HTML + Tailwind. No intermediate abstraction (no config files). The model writes what ships. Fewer layers = fewer tokens = faster output.

  3. MCP tools handle the heavy lifting. Image generation, product fetching, page publishing are all external tool calls. The model orchestrates them but doesn't do the compute.

Result: iteration speed matches your ad creative iteration. New ad? New landing page. Same effort.

What happens behind the scenes:

  1. Agent pulls your brand kit (colors, fonts, logo) via MCP

  2. Fetches product data from Shopify (real images, real price, real variants)

  3. Generates a hero background image matching your brand aesthetic

  4. Writes full HTML with Tailwind CSS, conversion-optimized section ordering

  5. Adds working commerce islands (BuyBox with add-to-cart, CartDrawer, ReviewCarousel)

  6. Validates page structure (no broken islands, no security violations, proper heading hierarchy)

  7. Publishes to your Shopify store with your permission

You get a live URL back. Page is functional immediately.

https://storefront.trylexsis.com/v/moxie-dandruff-detox-brez-style-pdp-v2?shop=moxiebeauty-jctnyrda.myshopify.com&preview=1

How Skills and MCP Work Together

You can use skills without the MCP server. You'll get excellent design guidance and HTML output, but you won't be able to generate images or publish directly to Shopify. For the full loop (prompt to live page), you need both.

Contributing

MIT licensed. The repo structure makes it straightforward to add packs:

your-pack-name/
├── claude/skills/     # One SKILL.md per knowledge area
├── claude/commands/   # Slash commands for specific workflows
├── codex/skills/      # Same content, Codex format
├── cursor/rules/      # Same content, .mdc rule files
└── install.sh         # Auto-installer

Useful contributions: new verticals (pet, fitness, electronics, SaaS), new traffic sources (Pinterest, email, affiliate), improved conversion patterns with data backing them.

Repo:

https://github.com/Aetheria-Labs1/storefront-skill-packs

Pages We Designed in Minutes

A few examples across different brands and verticals - each built from a single prompt:

  1. Cheat Day Powder Dry Shampoo | Moxie Beauty

  2. Moxie Beauty - Polished Euphoria-Inspired Storefront

  3. Khalifa Mints PDP | Khalifa Kush

  4. Moxie Dandruff Detox PDP - Brez Style

Different brands, different aesthetics, different product types. Same workflow. Same 60-second generation time.

To learn more about what we do:

https://trylexsis.com?utm_source=twitter&utm_medium=social&utm_campaign=storefront_skill_packs_launch&utm_content=about

What's Coming Next

More workflow integrations. Step-by-step guides for using skill packs with Claude Code, Codex, Cursor. We want any Shopify brand to be able to dynamically A/B test landing pages without touching code.

Per-user dynamic landing pages. Identifying each visitor uniquely and serving them a different page variant based on their intent signal - what they searched, what ad they clicked, what they browsed before.

Intent-based pages that rank. Landing pages built around search intent that are fully optimized for SEO and AI search engines (Google AI Overviews, Perplexity, ChatGPT). Pages that both convert paid traffic AND earn organic visibility through AEO compatibility.

Custom attribution tracking. Server-to-server attribution connected to your ad platforms - Meta CAPI, TikTok Events API. Know exactly which landing page variant drove which purchase, tied back to the original ad creative.

Closing

Five minutes to set up. Sixty seconds per page.

Star the repo:

https://github.com/Aetheria-Labs1/storefront-skill-packs

Sign up:

https://app.trylexsis.com/sign-up?utm_source=twitter&utm_medium=social&utm_campaign=storefront_skill_packs_launch&utm_content=article

Want us to do this for your brand? Book a demo:

https://trylexsis.com/?utm_source=twitter&utm_medium=social&utm_campaign=storefront_skill_packs_launch&utm_content=book_demo

Personalized storefronts, zero dev time.

Every ad, email, and AI agent gets its own landing experience. Powered by real customer signals from 40+ sources.

Related Articles

Success Case Study on DRANKS

CASE STUDY

Case Study on how Lexsis AI helped a new Indian beverage brand DRANKS - go from zero AI visibility to appearing in ChatGPT, Perplexity, and Google AI Overviews, competing against Shark Tank-funded competitors without ads or investor PR

Read

AI STOREFRONTS

Learn how personalized landing pages match every ad click to a unique page experience, lifting conversion rates by 30-80% without increasing ad spend.

Read
Personalized Landing Pages for Every Ad Click: How It Works

AI STOREFRONTS

97% of ad spend is wasted on generic landing pages. Learn how personalized post-click experiences lift CVR by 30-200% and turn paid traffic profitable.

Read