The Prompt That Changed My Mind
I typed this into Claude Desktop:
"Find me the best price for an RTX 4090 across Singapore, China, and US markets. Give me a side-by-side comparison."
Claude thought for a second, then did something I did not expect — it reached across borders, queried live inventory across 3 countries, and returned a ranked table.
No browser automation. No scraping. No "I cannot browse the internet."
Here is the MCP server that made it possible.
The @buywhere/mcp-server Package
export BUYWHERE_API_KEY=bw_live_xxxx
npx -y @buywhere/mcp-server
That is it. 30 seconds to install. The server exposes tools for:
-
search_products— Query across SG, CN, US, JP, KR, AU markets -
get_product_details— Pricing, availability, promotions -
compare_prices— Side-by-side across retailers like Lazada, Shopee, Amazon
Why This Matters for AI Agents
Agents are great at generating text. They are terrible at purchasing physical goods. The disconnect is that every commerce platform has a different API (if they have one at all).
MCP solves this. One protocol. One schema. One npx command.
The A2A Agent Card
BuyWhere also exposes an Agent-to-Agent discovery card at buywhere.ai/.well-known/agent.json. This means AI agents can autonomously discover BuyWhere without being told about it.
Try It Yourself
- Get a free API key at https://buywhere.ai/api-keys
- Connect it to Claude Desktop, Cursor, VS Code, or any MCP client
- Ask your agent to find and compare products
What I Learned
Cross-border e-commerce APIs are fragmented, but MCP makes them composable. The same agent that writes your code can now negotiate international supply chains.
That is wild.
Built with TypeScript, MCP SDK, and multi-market product aggregation. MIT licensed.
Top comments (0)