The same yoga mat sells for $19.99 on Amazon and $34.99 on Walmart. Finding these gaps manually means checking products one by one across platforms — opening tabs, searching by category, comparing prices, and recording each find in a spreadsheet. A seller might catch two or three obvious gaps in an afternoon. The other fifty opportunities, buried across product categories and marketplace combinations, go unseen.
Cross-platform arbitrage is not new. Sellers have been buying low on one marketplace and selling higher on another for as long as multiple platforms have coexisted. What has changed is the speed and scale at which the gaps can be discovered. An MCP-connected AI agent running marketplace-wide scans can surface arbitrage signals across hundreds of product categories in the time it takes to manually check five products on two sites.
Why Price Gaps Exist — and Why They Persist
Price gaps between marketplaces are not anomalies. They are the natural result of fragmented platform ecosystems. A product listed on Amazon US for $24.99 might sell for $39.99 on Walmart because Walmart’s category has fewer competing listings, a different customer demographic, and less aggressive algorithmic repricing. The same product on Amazon UK might land at a different price entirely, shaped by local competition, VAT treatment, and fulfillment cost structure.
These gaps persist because most sellers operate inside a single platform. An Amazon seller sources products, optimizes listings, and competes against other Amazon sellers — rarely checking what the same product fetches elsewhere. Even sellers active on multiple platforms manage each one independently, with separate inventory, pricing strategies, and data. The cross-platform comparison that would reveal a pricing opportunity simply does not happen.
The result is a market inefficiency that rewards whoever looks first.
Manual Discovery Does Not Scale
Consider the arithmetic of manual arbitrage discovery. A seller picks a category — kitchen utensils — and opens Amazon, Walmart, and eBay. For each product on Amazon, the seller searches the same product on Walmart, checks whether a matching listing exists, notes the price, and compares. If the gap is large enough to cover fulfillment cost and target margin, the product goes on a shortlist.
At ninety seconds per product, the seller covers roughly forty products per hour. Across three marketplace pairs (Amazon-to-Walmart, Amazon-to-eBay, Walmart-to-eBay), that is thirteen cross-platform comparisons per hour. Multiplied across ten product categories and thousands of ASINs, the coverage rate is under one percent. The seller is making decisions based on what happened to be checked, not what the data actually contains.
This is not a judgment problem. It is a coverage problem. The seller knows how to evaluate an arbitrage opportunity. What the seller lacks is a systematic way to surface the candidates.
An AI Agent That Scans Cross-Platform Prices Systematically
The sorftime-seller-agent is an open-source MCP server that connects any MCP-compatible AI agent to structured marketplace data across over forty platforms. Instead of checking products one by one, a seller asks the AI agent to scan a category across marketplace pairs and return price gaps that exceed a minimum threshold.
Here is what a cross-platform arbitrage scan looks like in practice. The seller is investigating the kitchen utensils category on Amazon US, cross-referencing against Walmart US:
Scan the top 200 products in the kitchen utensils category on Amazon US.
For each product, check whether the same or a similar listing exists on Walmart.
Return every product where the Walmart price exceeds the Amazon price by at least 30%,
after accounting for estimated Walmart fulfillment costs.
For each match, show:
- Amazon ASIN and price
- Walmart item ID and price
- The raw price gap in dollars and percentage
- A note on whether the Walmart listing is from a third-party seller or Walmart itself
Example output: cross-platform price gap between Amazon and Walmart.
The AI agent calls the relevant MCP tools — product search by category on Amazon, product lookup on Walmart, price extraction, and gap calculation — and returns a structured table. What took a seller an afternoon of manual searching now takes under a minute. The coverage is not forty products. It is the full category, checked systematically, with no product skipped because the seller got tired or lost focus.
The Code: A Repeatable Cross-Platform Scan
The real leverage comes from composing the scan into a repeatable script. Here is an example using sorftime-seller-agent MCP tools, called from within an AI agent session. The tools are auto-discovered after installation:
#!/bin/bash
# Cross-platform price gap scan — Amazon US vs Walmart US
# Requires: sorftime-seller-agent installed and MCP agent connected
CATEGORY="kitchen-utensils"
GAP_THRESHOLD=30
MARKETPLACE_AMZ="US"
MARKETPLACE_WMT="US"
echo "=== Scanning ${CATEGORY} for cross-platform price gaps ==="
echo "Threshold: ${GAP_THRESHOLD}% minimum price difference"
echo ""
# Pull top products from Amazon category
sorftime product category-top \
--marketplace "$MARKETPLACE_AMZ" \
--category "$CATEGORY" \
--limit 200 \
--fields asin,title,price \
| jq -r '.data[] | "\(.asin)\t\(.title)\t\(.price)"' \
| while IFS=$'\t' read -r asin title amz_price; do
# Search Walmart for matching product
wmt_result=$(sorftime product walmart-search \
--marketplace "$MARKETPLACE_WMT" \
--query "$title" \
--limit 1 \
| jq -r '.data[0] | "\(.item_id)\t\(.price)\t\(.seller_type)"' 2>/dev/null)
if [ -n "$wmt_result" ]; then
IFS=$'\t' read -r wmt_id wmt_price seller_type <<< "$wmt_result"
# Calculate gap
gap=$(echo "scale=1; ($wmt_price - $amz_price) / $amz_price * 100" | bc)
if (( $(echo "$gap >= $GAP_THRESHOLD" | bc -l) )); then
echo "[GAP] $title"
echo " Amazon: \$${amz_price} | Walmart: \$${wmt_price} | +${gap}% | Seller: ${seller_type}"
echo " ASIN: ${asin} | Walmart ID: ${wmt_id}"
echo ""
fi
fi
done
A seller runs this script against a category and gets a ranked list of arbitrage candidates — with prices, gap percentages, and seller types — in under sixty seconds. The same script modified to target a different marketplace pair (Amazon Japan to Amazon US, Walmart to eBay) follows the same pattern. The tool handles the data plumbing. The seller handles the decision.
What the Numbers Actually Mean
A price gap alone is not an arbitrage opportunity. The gap must exceed the cost of moving a product from one platform to the other. That cost includes marketplace referral fees on the selling platform, fulfillment costs (which differ between FBA, WFS, and self-fulfillment), shipping between warehouses, returns processing, and category-specific fees. A 30% gap that shrinks to 8% after fees is not an opportunity. A 45% gap that holds at 22% after fees is.
The AI agent can layer fee estimation into the same scan. A seller asks for net arbitrage margin — price gap minus estimated fulfillment and referral costs on the target platform — and the agent filters to only the candidates that clear the seller’s minimum margin threshold. The data points are the same. The arithmetic that used to require a separate spreadsheet per candidate now runs in the same query.
Integrating Arbitrage Scanning Into a Weekly Workflow
The scan is most useful when it runs regularly. Marketplaces reprice constantly. A gap that exists on Tuesday morning may close by Tuesday evening. A script that runs daily or weekly and flags only new or widening gaps gives a seller a steady feed of actionable leads without requiring active hunting.
A seller can extend the scan across additional marketplace pairs — Amazon US to Amazon UK, Walmart to eBay, Amazon Japan to Amazon US — and across an expanding set of categories. The underlying tool is the same. The only variable is which marketplaces and categories the seller cares about.
This is not about automating the seller’s judgment. It is about automating the data collection that precedes judgment. The seller still evaluates each candidate: Is the product gated on the target platform? Is demand sufficient to absorb additional supply? Are there brand restrictions or IP considerations? The tools deliver the signal. The seller applies the context.
Limitations Worth Knowing
Cross-platform arbitrage carries real constraints that a price gap alone cannot surface. Brand-gated categories on Walmart and Amazon prevent unauthorized resellers from listing certain products. Some platforms enforce MAP pricing agreements that cap the selling price regardless of what the market would bear. Fulfillment costs vary by product dimensions and warehouse location in ways that a simple percentage estimate can miss. And a gap that looks wide today may narrow when competitors with the same tooling act on the same signal.
The sorftime-seller-agent identifies price discrepancies. It does not guarantee that a given discrepancy is executable as a business. The tool reduces the discovery cost to near zero. The due diligence remains the seller’s responsibility.
Getting Started
The tool installs with a single command and connects to any MCP-compatible AI agent:
git clone https://github.com/DannylydST/sorftime-seller-agent
cd sorftime-seller-agent
python3 scripts/install.py
Free API access at open-intl.sorftime.com. After installation, restart your AI agent and ask it to scan a category across marketplace pairs. The tools are auto-discovered. The data is live. The next arbitrage opportunity is a query away.