If you build AI agents that need real Japanese data, you've probably hit the same wall I did: Japanese e-commerce, price indices, government datasets, and local marketplaces are scattered across sites with no clean API. So over the past few months I built and published 10 MCP (Model Context Protocol) servers that wrap these sources into tools any MCP client (Claude Desktop, Cursor, Cline, etc.) can call directly. All are free to connect, and the source is on GitHub.

The list

Server What it gives your agent Smithery GitHub
japan-ec-mcp Japanese e-commerce product/price lookups (Mercari, Rakuten, Yahoo! Auctions) connect repo
japan-market-mcp Japanese market & price-index snapshots connect repo
japan-fuel-price-mcp Weekly official fuel prices across Japan connect repo
tcg-price-japan Trading-card (Pokemon etc.) prices from Japanese markets connect repo
japan-anime-figure-mcp Anime figure prices & availability in Japan connect repo
kensho-sweep-mcp Live Japanese sweepstakes/giveaway listings (X-based) connect repo
kensho-kclub Sweepstakes aggregator data (kenshou.club) connect repo
kensho-kema Sweepstakes aggregator data (ken-kaku / ke-ma sources) connect repo
kensho-kaku Kakaku.com-style price comparison data connect repo
japan-ec-apify-mcp EC data routed through managed Apify actors (no scraping setup) connect repo

Official MCP Registry entry

One of them — mlit-property-prices-mcp — is also published on the official MCP Registry as io.github.atushi1841/mlit-property-prices-mcp (status: active). It exposes Japan's MLIT land-transaction price dataset (XIT001) — the actual recorded sale prices of real estate parcels — which is otherwise painful to query programmatically. If you're building anything real-estate or investment related for the Japanese market, this is the primary-source data behind the headlines.

How to connect

Every server is hosted on Smithery, so connecting is one URL:

{ "mcpServers": { "japan-ec-mcp": { "command": "npx", "args": ["-y", "@smithery/cli@latest", "run", "atushi1841/japan-ec-mcp"] } } } 

Or just open the Smithery page for any server above and copy its connection snippet into Claude Desktop / Cursor config.

Why I built these

I run a personal automation project around Japanese sweepstakes and price monitoring, and every data source I needed was HTML-only. Rather than one-off scrapers, I packaged each domain as an MCP server so my agents (and eventually yours) can call them as plain tools. The code is on GitHub under the atushi1841 namespace — MIT-style, no API keys required for the free endpoints.

What I'd like feedback on

  • Which data domain would be most useful as a hosted MCP tool? (EC prices, government datasets, and sweepstakes are the current three families.)
  • Are there MCP clients where Smithery-hosted servers fail to connect cleanly? I want to fix transport issues.

If any of these save you a scraping weekend, a bookmark or a star on the repo genuinely helps these show up for other people building Japan-facing agents.