japan-travel-mcp-data

The runtime data for the japan-travel-mcp Model Context Protocol server. Comprehensive Japanese travel data for AI agents, built from public official sources, covering all 47 prefectures and 1,938 local government entities.

Code lives on GitHub: github.com/ookami0210/japan-travel-mcp Data lives here. The npm package downloads this dataset on first run.

Why this dataset exists

Japan's tourism information — created to reach the world — is published across thousands of municipal websites. Almost none of it is accessible to AI agents in a structured, multilingual form. This dataset fixes that gap.

— KJ Sunada, founder of KabuK Style

What's inside

translations/
  descriptions_complete.jsonl     # 13,394 attractions × 18 languages — rich
                                  # 200-300 char tourism descriptions
  multilingual_complete.jsonl     # 13,961 attractions × 18 languages — names
  multilingual_wikipedia.jsonl    # 18-language names from Wikipedia sitelinks
  jp_en.jsonl                     # JP → EN canonical name mapping

prefectures/                      # 47 prefecture files: municipal-scrape spots
                                  # + Wikidata attractions per prefecture

hotels/
  master.json                     # ~20,000 accommodations (Wikidata + OSM merged)

r3/                               # Official designation registries
  maff_gi.json                    # 172 MAFF Geographical Indications (food / agri-products)
  meti_densan.json                # 231 METI-designated Traditional Crafts (Dentō Kōgeihin)
  japan_heritage.json             # 104 Japan Heritage stories (Nihon Isan)
  bunka_intangible.json           # 125 Important Intangible Cultural Properties
  unesco_japan.json               # 58 UNESCO ICH inscriptions for Japan
  translations/
    r3_translations.jsonl         # 690 designation records × 18 languages

glossary/
  seed_canonical.json             # House style for translations
  mlit_canonical.json             # Japan Tourism Agency (MLIT) official terminology
  wikipedia_multilingual.json     # 18-language Wikipedia sitelinks (build-time)

_state/
  wikidata_attractions.json       # 41,404 Wikidata attractions, ja-anchored
  municipalities.json             # 1,938 municipalities + designated-city wards
  municipality_centroids.json     # JIS-coded centroid coordinates
  official_urls.json              # Resolved official tourism site URLs

Source policy — official build-up only

This dataset only contains records that an authoritative public body has designated, scraped from that body's own publication. No editorial picks, no AI-curated lists, no UGC.

17 supported languages

English (en), Japanese (ja), Chinese Simplified (zh), Korean (ko), French (fr), Spanish (es), German (de), Italian (it), Portuguese (pt), Russian (ru), Thai (th), Vietnamese (vi), Indonesian (id), Malay (ms), Arabic (ar), Hindi (hi), Tagalog (tl).

The 17 were chosen to cover the JNTO inbound-tourism priority languages plus major source markets across Asia-Pacific, Europe, and the Middle East.

Coverage

All 47 prefectures are populated; every entity has descriptions in all 17 languages (no per-language gaps inside the 13,394-entity description set). The chart shows the per-prefecture entity count — the long tail outside Kyoto / Tokyo / Hokkaido is the actual point of this dataset.

13,394 attractions × 18 languages = 241,092 description cells. Traditional Chinese (zh-Hant, Taiwan lexicon) is derived from the quality-controlled Simplified layer via deterministic OpenCC conversion (s2twp); records added after 2026-08 are generated natively in both Chinese scripts. Plus 690 official-designation records (MAFF GI, METI crafts, Japan Heritage, Bunka-cho intangible records, UNESCO ICH) translated to the same 17 languages = 12,420 more cells. Plus 13,961 canonical names × 18 languages = 237,337 more cells.

Refresh cadence

The GitHub Actions cron in the code repo refreshes data on two tracks and re-publishes to this dataset:

Each domain is hit at most once per cycle.

How to use

Via the MCP server (recommended)

npm install -g japan-travel-mcp
japan-travel-mcp                  # downloads this dataset to ~/.japan-travel-mcp/data/ on first run

Then add to your AI agent's MCP config (Claude Desktop, Cursor, etc.).

Direct download

from huggingface_hub import snapshot_download
local_dir = snapshot_download(
    repo_id="open-travel/japan-travel-mcp-data",
    repo_type="dataset",
)

# Or via git-lfs:
git clone https://huggingface.co/datasets/open-travel/japan-travel-mcp-data

Citation

If you use this dataset in research or a product, please cite:

KJ Sunada, "Japan Travel MCP", 2026.
GitHub: https://github.com/ookami0210/japan-travel-mcp
HF dataset: https://huggingface.co/datasets/open-travel/japan-travel-mcp-data
License: CC BY 4.0

License

Data: CC BY 4.0 — free to use including commercially, attribution required.

Code (separate repo): MIT.

Underlying source data carries its own licenses (see source-policy table above).