Connect MLIT Data Platform to Claude

This server connects your AI client to the MLIT Data Platform user API (国土交通DPF利用者API), operated by Japan's Ministry of Land, Infrastructure, Transport and Tourism. It aggregates infrastructure and geospatial datasets from across the ministry and beyond behind a single GraphQL endpoint.

The tools form a find -> detail -> file chain:

• list_data_catalogs shows which catalogs and datasets exist (rsdb = nationwide road facility inspections, cals = public works and survey records, ngi = ground survey data, nlni_ksj = national land numerical information, mlit_plateau = PLATEAU 3D city models, rtc = road traffic census, hwq = water quality, and more).
• search_data searches by Japanese keyword and/or catalog_id / dataset_id, with optional bounding-box or radius filters and sorting. When a keyword returns nothing, suggest_keywords surfaces terms that actually occur in the data (e.g. 川 -> 河川, 河川敷) with hit counts.
• get_data returns one item's full detail (every metadata attribute, which vary per dataset, plus attached files and licence), and get_file_download_urls turns those files into download URLs for drawings, reports and survey data (URLs issued by DPF expire after 60 seconds).
• get_area_codes resolves prefecture and municipality codes, with kanji, katakana and romaji names plus validity dates so merged municipalities can be traced.

Arlet obtains and manages the API key, so there is nothing for you to apply for or configure - just switch the server on. (Under the upstream terms of use, heavy bursts of requests on a single key may be throttled.)

Notes: dataset licences differ per catalog, so check the licence of each catalog you rely on (list_data_catalogs with include_datasets=true returns them). Services built on this API must display the credit: 「このサービスは、国土交通データプラットフォームのAPI機能を使用していますが、最新のデータを保証するものではありません。」 Search paging reaches the first 10,000 hits.

If you need the 3D city models themselves (per-building features and attributes, CityGML, the specification), the dedicated PLATEAU server is a better fit.