daraz-mcp

Daraz MCP Server - Clean & Fast

A streamlined Model Context Protocol (MCP) server for searching Daraz.pk products. Built using documented Daraz API patterns for maximum reliability and performance.

✨ Features

• 🎯 Simple & Clean: Just 2 tools instead of 6 confusing ones
• 📡 Official API Patterns: Uses documented mods.listItems structure and ajax=true endpoints
• 💸 Smart Cheapest Search: Auto-detects "cheapest" queries and searches extensively
• 🚫 No Cache: Always fresh results, no .db files created
• 🔄 Multi-method: JSON API primary, Playwright browser fallback
• 🎨 Beautiful Output: Human-readable results with clickable links
• ⚡ Fast: Optimized using officially documented endpoints

🛠️ Tools Available

1. search_daraz - The Main Search Tool

Smart search that handles everything automatically:

# Regular search
search_daraz("wireless mouse")

# Auto-detects cheapest requests  
search_daraz("cheapest wireless mouse")  # Searches 15 pages, sorts by price

# Explicit cheapest mode
search_daraz("mouse", cheapest=True)

# Price filtering
search_daraz("laptops", max_price=50000)

# Category search (uses documented category endpoints)
search_daraz("TV", category="televisions")

Features:

• Auto-detects "cheapest" in query and switches to extensive search mode
• Searches 15 pages for cheapest queries vs 5 for regular
• Sorts results by price for cheapest queries
• Returns formatted, clickable results

2. product_details - Get Product Info

product_details("https://www.daraz.pk/products/product-url")

📡 API Implementation

This server uses officially documented Daraz.pk endpoints:

Search Endpoint

GET https://www.daraz.pk/catalog/
    ?ajax=true
    &q=query
    &page=1
    &_keyori=ss

Category Endpoint

GET https://www.daraz.pk/{category}/
    ?ajax=true
    &page=1
    &q=search_within_category

Response Structure: data.mods.listItems[] contains product array

🚀 Quick Setup

1. Install dependencies:

pip install fastmcp requests beautifulsoup4 playwright

1. Add to LM Studio mcp.json:

{
  "mcpServers": {
    "daraz-search": {
      "command": "C:\\path\\to\\python.exe",
      "args": ["C:\\path\\to\\server_merged.py"]
    }
  }
}

1. Test in LM Studio:

• "Find cheapest Nothing phone case"
• "Search for wireless mice under 2000 PKR"
• "Show me gaming laptops"

🎯 Improvements from Previous Version

Before (Confusing):

❌ 6 different tools: search_daraz, search_daraz_structured, search_daraz_formatted, search_cheapest_daraz, search_most_expensive_daraz, product_details ❌ Cache causing stale results ❌ AI couldn't decide which tool to use

After (Clean):

✅ 2 simple tools: search_daraz (smart), product_details ✅ No cache - always fresh results ✅ Auto-detection of user intent ✅ Uses documented API patterns

📊 Search Behavior

| Query Type | Pages Searched | Results | Sorting | |------------|---------------|---------|---------| | Regular search | 5 pages | Up to 10 | As found | | "Cheapest" query | 15 pages | 1 item | Price (lowest first) | | With cheapest=True | 15 pages | 1 item | Price (lowest first) | | With max_price | 5 pages | Filtered | As found |

🔧 No Cache = Always Fresh

• ✅ No .db files created
• ✅ Every search hits Daraz API fresh
• ✅ Latest prices and stock status
• ✅ No stale cached results

🛡️ Anti-Bot Protection

• User-Agent rotation (5 different browsers)
• Random delays between requests (1.0-2.5s)
• Proper headers and referrers
• Graceful fallback to browser automation

📖 Usage Examples

For Users in LM Studio:

• "Find me the cheapest Nothing phone 1 case"
• "Search for wireless mice under 1500 PKR"
• "What are some good gaming laptops?"
• "Show me TVs in the television category"

For Developers:

# The AI will automatically choose the right parameters
search_daraz("cheapest phone cases")  # cheapest=True, max_results=1
search_daraz("phone cases")           # regular search, max_results=10
search_daraz("cases", max_price=500)  # price filtering

🎉 Ready to Use

This server is production-ready and uses officially documented Daraz.pk API patterns for maximum reliability. No seller account required - works with public endpoints only.

Start LM Studio, add the MCP server, and start shopping! 🛍️