Product Recommendation & Quality Scoring

AI-powered product evaluation, similarity comparison, and alternative recommendations for cross-border e-commerce.

When to Use

• User asks “is this product any good?”
• User wants to compare two similar products
• User asks for alternative or similar products
• User has a list of products and wants them ranked
• User asks about product quality signals on Chinese platforms
• User needs help deciding which product to buy

MCP Server Setup

{
  "mcpServers": {
    "product-recommendation": {
      "command": "npx",
      "args": ["-y", "@shopme/product-recommendation-mcp"],
      "env": {
        "GEMINI_API_KEY": "your-gemini-key"
      }
    }
  }
}

GEMINI_API_KEY is only required for recommend_alternatives. Other tools work without it.

Available Tools

score_product

Evaluate a product on multiple quality dimensions.

• Input: Product data (name, price, rating, soldCount, reviewCount, shopRating, platform)
• Optional: averagePrice for value comparison
• Returns: Overall score (0-100), dimension scores, analysis, verdict

find_similar

Compare two products for similarity.

• Input: Two product objects
• Returns: Similarity score (0-1), matched features, differences, recommendation

recommend_alternatives

Get AI-powered alternative product suggestions.

• Input: Product name, optional category/budget/preferences
• Returns: 3-5 alternative suggestions with search keywords and platforms
• Requires: GEMINI_API_KEY

rank_products

Batch score and rank multiple products.

• Input: Array of products (2-20)
• Returns: Ranked list with scores and top pick

Scoring Dimensions Explained

Score Interpretation

Quality Signals on Chinese E-commerce

Positive Signals

• Rating ≥ 4.8 with 1000+ reviews → Highly trustworthy
• Sold 10,000+ units → Battle-tested product
• Shop rating ≥ 4.8 → Reliable seller
• Multiple detailed images → Professional listing
• Tmall brand store → Official/authorized

Red Flags

• No reviews → Too new or fake listing
• Rating < 3.5 → Known issues
• Price is >50% below average → Possible quality issues or counterfeit
• Stock photos only → May not match actual product
• New shop with 0 ratings → Higher risk

Recommendation Workflow

For a single product decision:

1. score_product → Get quality assessment
2. If score < 65, use recommend_alternatives → Get better options
3. Search for alternatives using suggested keywords

For choosing between options:

1. rank_products with all candidates → Get ranked list
2. find_similar on top 2 → Understand key differences
3. Choose based on your priorities (price vs quality vs shipping)