AgentShelf

User Guide

AgentShelf audits your public storefront for AI readability and generates a structured llms.txt file that AI agents like ChatGPT, Claude, Perplexity, and Gemini can understand.

What is AgentShelf?

AgentShelf scans your store's public pages, detects your platform (WordPress, Shopify, Etsy, WooCommerce, Strapi, Contentful, or generic), and scores how well AI agents can read your catalog.

Key Features

  • Platform Detection — Automatically identifies your CMS or commerce platform
  • AI Readability Score — Transparent 0–100 breakdown showing what agents see
  • llms.txt Generation — Structured format for AI agents to cite your products
  • MCP Tools — First-class integration with Cursor and Claude Desktop
  • Multi-Platform Support — Works with 7+ platforms out of the box

How It Works

1. Submit Your Store URL

Paste your store's homepage URL into the audit form. AgentShelf fetches the page with a polite bot user-agent and respects size/timeout limits.

2. Get Your Score

Receive a 0–100 AI readability score based on:

  • Platform signals — Recognized CMS/commerce platform
  • Product structure — JSON-LD, microdata, or HTML product cards
  • Metadata quality — Open Graph, titles, descriptions
  • Content accessibility — Headings, lists, semantic markup

3. Download llms.txt

Your llms.txt file includes:

  • Site title and description (H1 + summary blockquote)
  • Product catalog section with linked entries
  • Platform-specific notes on data freshness
  • Proper markdown formatting for agent parsing

Supported Platforms

AgentShelf works with:

WordPress & WooCommerce

  • WP REST API when public (/wp-json/wp/v2/posts)
  • HTML fallback for locked REST endpoints
  • WooCommerce product loop detection

Shopify

  • Public /products.json and /collections.json
  • No Admin API required
  • Collection-based sections in llms.txt

Etsy

  • Public shop HTML parsing
  • Listing cards → structured product previews
  • OAuth support planned for future release

Strapi

  • Auto-detection via page metadata
  • Public /api/<type> endpoints
  • Optional API token via environment variable

Contentful

  • CDN/meta detection
  • Delivery API with space ID + access token
  • Credentials via environment variables

Generic CMS

  • JSON-LD and Open Graph parsing
  • HTML product card extraction
  • Fallback for any public website

Understanding Your Score

Your score ranges from 0 to 100:

  • 90–100 — Excellent. Structured data, clear product catalog, agent-ready.
  • 70–89 — Good. Most data is accessible; minor improvements suggested.
  • 50–69 — Fair. Agents can read your site but may miss products.
  • Below 50 — Poor. Significant improvements needed for AI readability.

Breakdown Components

Each audit shows:

  • Platform detected — Confidence and identification method
  • Products found — Count and quality of product data
  • Metadata — Open Graph, titles, descriptions
  • Structure — Semantic HTML, JSON-LD, headings
  • Recommendations — Specific improvements for your platform

Using MCP with Cursor / Claude

AgentShelf includes Model Context Protocol (MCP) tools for direct AI agent integration.

→ MCP Setup Guide

Frequently Asked Questions

Is AgentShelf free?

Yes! Public URL audits, MCP tools, and llms.txt generation are free forever.

AgentShelf Pro (coming soon) will add:

  • Bot analytics and traffic tracking
  • Scheduled catalog re-sync
  • One-click publish for Shopify, WordPress, and Etsy

Do you store my products?

No. AgentShelf fetches your public pages on-demand and does not store product data. Each audit is stateless.

Can I use this for private/staging stores?

Currently, AgentShelf only audits public URLs. Private or authenticated stores are not supported.

What about my API credentials?

For platforms that require credentials (Strapi, Contentful), you provide them via:

  • MCP tool arguments (never logged)
  • Environment variables (local or Cursor/Claude config)

AgentShelf never stores or logs your tokens.

How often should I re-audit?

Re-audit whenever you:

  • Add or remove products
  • Change your platform or theme
  • Update meta descriptions or structured data
  • Want to verify llms.txt freshness

AgentShelf Pro will support scheduled re-audits.

Can AI agents actually use llms.txt?

Yes! llms.txt is a lightweight markdown format that:

  • ChatGPT, Claude, Perplexity, and Gemini can fetch and parse
  • Provides structured product listings with links
  • Works as a "sitemap for AI agents"

Place your llms.txt at your domain root (e.g., example.com/llms.txt) for best results.

Next: MCP Setup → or Platform Details →