Connect Product Data to Your AI Agent with MCP

Name: ShelfRun
Author: ShelfRun

March 1, 2026 · 6 min read

The Model Context Protocol (MCP) lets AI agents use external tools. ShelfRun's MCP server gives your agent access to real-time e-commerce data — product prices, stock levels, and competitive intelligence.

What is MCP?

MCP is an open protocol that lets AI models (Claude, GPT, etc.) call external tools. Instead of hardcoding API calls, your agent discovers available tools and uses them naturally through conversation.

Install ShelfRun MCP Server

npx @shelfrun/mcp

Configure in Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "shelfrun": {
      "command": "npx",
      "args": ["@shelfrun/mcp"],
      "env": {
        "SHELFRUN_API_KEY": "sr_live_your_key"
      }
    }
  }
}

Available Tools

Your agent gets access to 6 tools:

scan_store — Detect platform and capabilities for any URL
scrape_products — Extract all products from a store
get_products — Get products from a saved store
list_stores — See all monitored stores
add_store — Add a new store to monitor
compare_prices — Compare products across stores

Example Conversation

You: "What are the current prices on allbirds.com?"

Claude: *uses scan_store("allbirds.com")*
        *uses scrape_products("allbirds.com")*

"Here are the current Allbirds prices:
- Tree Runner: $98.00
- Wool Runner: $110.00
- Tree Dasher 2: $135.00
...247 products found in total."

Why MCP for E-Commerce Data?

Real-time: Agent gets live data, not stale training data
Natural language: Ask questions like “which store has the cheapest running shoes?”
No code needed: The agent handles all the API calls
Composable: Combine with other MCP tools for powerful workflows

Build Autonomous Agents

Combine ShelfRun MCP with other tools to build agents that:

Monitor prices and update a Notion database
Find the cheapest source for a product across 10 stores
Generate weekly competitive intelligence reports
Alert your team on Slack when a competitor changes pricing