Japan is live on Croma, our eighth country and the first outside the Americas. It arrives with two sources: the register of corporate numbers the National Tax Agency keeps, and the case law the courts publish. Four new endpoints behind the key you already use, every one also an MCP tool, so an agent can call them without you writing a client.
Every registered entity in Japan#
The National Tax Agency assigns every company, foreign company, government body and other registered entity a 13-digit corporate number (法人番号) and publishes the register behind it: about six million records. Each one comes with the registered name, its reading in kana and its English form where the company registered one, the legal form (株式会社, 合同会社, 一般社団法人 and the rest), the head office down to the municipality code and postal code, the status of the registration with the date and reason of any closure, the successor after a merger, and the last change the register recorded.
Search by any part of the name in kanji, kana or English, so トヨタ finds トヨタ自動車株式会社 and every company with トヨタ in its name, and narrow by prefecture, city, legal form, status, or the kind and date of the last change: change_type: "new" with a date range lists the companies registered in a period. Or look one company up by its number, which the endpoint checks before asking.
The courts' case law#
About 68,000 rulings the courts of Japan publish since 1947: the Supreme Court's judgments and decisions, the high courts' reports, the lower courts' bulletin and the collections of administrative, labour and intellectual property cases. Each ruling comes with:
- its case number and name, the court and bench, the date and the kind of decision,
- the outcome, the lower court's decision, and the official reports citation where it is reported,
- the court's own summary where it writes one (判示事項, 裁判要旨) and the statutes it cites,
- the full text, read from the court's document.
Search in Japanese across every collection at once: free text reaches into the summaries and the body of every ruling, and collection, court, case_number, year and a date range narrow it.
| Endpoint | What it answers |
|---|---|
nta-companies-search | Search the corporate number register |
nta-company | One company by its corporate number |
courts-jp-rulings-search | Search the courts' case law |
courts-jp-ruling | One ruling in full, with its text |
The original document, every time#
We keep a copy of every document the courts publish, byte for byte as served. document_url points to that copy, so if a link moves or disappears at the source, the document is still there. The register publishes an image for the one name or address in a hundred it cannot print in full; those are kept the same way.
An example#
Toyota, by name and prefecture:
curl https://api.croma.run/jp/nta/companies-search/v1 \
-H "Authorization: Bearer $CROMA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"query": "トヨタ自動車",
"prefecture": "愛知県"
}'Every result carries the corporate_number you pass to nta-company. Change the path to /jp/courts-jp/rulings-search/v1 and send {"query": "解雇権濫用", "collection": "supreme"} to find the Supreme Court's rulings on the abuse of the right to dismiss.
How current it is#
The register publishes its changes every business day and Croma takes them in the same evening, so a search is at most a business day behind the register. The courts' collection refreshes every day with what the courts posted. Every response carries as_of, the date the data is current to.
The full archive is loading this week: the register arrives prefecture by prefecture over a few hours, and the rulings back to 1947 over the next days, the most recent first.
Fields never carry the parties' names. The text stays as the courts published it.
The guides have every field: National Tax Agency and Courts of Japan. To call the endpoints from Claude, ChatGPT or Cursor, connect the MCP server.