Introduction

JMAD, the Japan Master Asset Database, is DNK's growing public database of buildings in Japan, with permanent IDs and source attribution. Coverage varies by area and field. Every value on every record carries its source, the URL it was pulled from, when it was retrieved, and how confident the extraction is. You can look up a building, read the number, and click through to the page it came from.

The base layer is free and keyless. No account or API key is required. Public request limits apply. curl an asset ID and you get JSON back.

Who reads this

• Asset managers and analysts joining portfolio data to a stable building identifier instead of a fuzzy address string.
• Due diligence teams checking a building's floor count, height, or ownership against more than one source before they rely on it.
• Mapping and research projects that need footprints, floor counts, and building types at country scale.
• AI agents and the people building them, reading structured data or markdown and citing a source URL instead of guessing.

How a fact gets onto a page

JMAD pulls the same building from several sources: OpenStreetMap, PLATEAU, the Address Base Registry, e-Stat census boundaries, Wikidata, Nominatim, J-REIT and EDINET disclosures, and the Tokyo green-building and 管理計画認定 (mankan) registers. Sources disagree. Wikidata might list a building at 179.2 m while OpenStreetMap has 18 m for the same field, both pulled from real records. JMAD picks a winner per field with a precedence table (OSM and Wikidata usually settle names and asset type; PLATEAU and OSM settle footprint; J-REIT and EDINET disclosures settle ownership) and keeps the losing values on the record instead of dropping them. Every page shows the winning value's provenance and any conflicting value it beat. Nothing is silently overwritten: a correction is a new record, and the old one stays in history. See How JMAD is built for the pipeline that produces this, and Provenance and precedence for the precedence table itself.

Three layers

The public layer doesn't move behind a wall later. It's the foundation the other two sit on, not a teaser for them.

Licensing

Each source carries its own licence: OpenStreetMap and Nominatim are ODbL, Wikidata is CC0, PLATEAU and e-Stat are CC BY 4.0, the Address Base Registry, EDINET, the Tokyo green-building label and the mankan register are government open data with dataset-specific terms, and J-REIT issuer sites and the Tokyo green-building archive are proprietary. Fields JMAD computes itself, like GFA from footprint and floor count, are marked as derived. See Licensing and attribution for what you need to carry when you reuse a fact.

Where to go next

• Quickstart to read a building in under a minute.
• How JMAD is built for the gather, save, consolidate, publish pipeline.
• DNK asset IDs for the identifier scheme every record is keyed on.
• Field dictionary for what every field means and where it typically comes from.
• Provenance and precedence and licensing and attribution for how conflicts get resolved and what you can carry when you reuse a fact.
• Authentication and API overview if you're building against the API directly.

Browse and connect

The homepage has a building search and a Build with this data panel. Expand Connect an agent with MCP for the server URL and connection configuration, or Use the API for a copyable request. The same controls appear on area lists and building records.

Building records organize their facts, sources, history, and API/MCP access into tabs. Recent changes shows recorded changes from recently published building records.