Japan Corporate Number MCP Server

@pipeworx/japan-corporate-number

Japan's National Tax Agency (国税庁) corporate-number register (法人番号公表サイト) — the complete spine of every 13-digit 法人番号 issued since 2015-10-05: legal name, katakana reading and English name when NTA holds one, address, corporate kind (stock company, LLC, government body, NPO, etc.) and status (active, or closed/merged with the narrative NTA published). 5,837,980 entities as of the 2026-09-30 cut.

Part of Pipeworx — an MCP gateway connecting AI agents to 1769+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

gBizINFO (mcps/gbizinfo, METI) is a different source hanging richer government records — procurement, subsidies, patents, financials — off the same corporate number, with partial coverage. This pack is the complete spine (every registered number, from NTA directly); use gbizinfo for the government-record detail once you have a number.

Tools

• jp_corporate_number(number) — look up one entity by its 13-digit 法人番号. Validates the check digit before looking up (see "Check digit" below) and returns found: false with a reason for a malformed number, a check-digit failure, or a correctly-formed number simply not in the register — never an empty object.
• jp_company_search(name, prefecture?, limit?) — search by Japanese name (substring, corporate-suffix-insensitive) or, for the subset of entities NTA published one for, an English name/katakana reading. prefecture accepts a Japanese name, an English name, or a 2-digit code.

Auth

Keyless. The gateway supplies the data directly — no caller-facing key, and no vendor key of any kind (the register is published as a monthly file, not a live API). The tools work only through the hosted gateway: a standalone build has no data to read and returns a loud error (StoreUnavailableError) naming that, never an empty result.

Data source & reuse terms

• Japan National Tax Agency (国税庁), Corporate Number Publication Site (法人番号公表サイト) — the whole-country corporate-number register, published monthly. Pipeworx loads each monthly edition; data_as_of on every response is when the loaded edition was taken. The fetch mechanics (and the verified column mapping) live only in scripts/japan-corporate-number/README.md, which does not publish — CLAUDE.md step 9b (a4c024647): name the source, never the source file/path/form fields.
• Terms of use (Japan's Public Data License v1.0, quoted from NTA's terms-of-use page): attribution is required ("本コンテンツを利用する際は出典を記載してください" — when using this content, state the source); edits/derivative processing must be disclosed separately from the attribution; and edited/processed data must never be presented as though the government itself produced it ("編集・加工等した情報を、あたかも国（又は府省等）が作成したかのような 態様で公表・利用することは禁止します"). Otherwise permissive — commercial reuse and redistribution are not restricted. This README and every tool response carry the source and data_as_of; the data is NTA's, reformatted by Pipeworx, never presented as government-produced.

Kind codes (法人種別)

The complete set observed: 101 national government organization, 201 local government, 301 Kabushiki Kaisha (stock company), 302 Godo Kaisha (LLC-equivalent), 303 Gomei Kaisha, 304 Goshi Kaisha, 305 other registered corporation (e.g. incorporated association/foundation), 399 other, 499 foreign company etc. Labels are best-effort English glosses of NTA's standard code table; the raw code is always returned alongside the label.

Status — a real correctness trap (and a second one inside it)

A corporation closed by merger (process_type 71/72, ~31,000 entities) carries closure_event_text (a Japanese sentence naming the other party) but closure_date and closure_cause_code are NULL for these rows — that pair is populated by a different closure mechanism entirely (mostly process_type 21, ~784K rows), which mostly has no closure_event_text. Checking closure_date alone misses every merged-away entity.

The trap inside the trap: closure_event_text describes a merger from either side, and only the wording says which. "...を合併" (absorbed [other]) names the SURVIVOR — this entity continues (e.g. Toyota Motor Corporation absorbing a subsidiary). "...に合併し解散" (merged INTO [other] and DISSOLVED) names the entity that disappeared. Treating any closure_event_text as closed is wrong — found live reporting an active, operating company as closed. status is closed_or_merged when closure_date is set, OR closure_event_text is set AND literally contains 解散 ("dissolved").

Check digit

A 法人番号's first digit is a check digit over the other 12 (most significant first), weighted 2,1,2,1,... (the most significant base digit carries weight 2), summed, mod 9, check = 9 - (sum mod 9). Verified against real numbers from the register (1000012160153, 1000013050386, 1000013050394 all pass; the naive "textbook" weighting 1,2,1,2,... fails all three). jp_corporate_number validates this before looking up, so a malformed or impossible number gets a specific bad_format / bad_check_digit reason rather than a bare "not found".

"Japanese and romanised matching" — what this is and is not

jp_company_search matches a single FTS5 trigram index over the Japanese core name (corporate-suffix stripped, e.g. "株式会社" / "K.K." both fold away), the English name, and the katakana reading — whichever of these NTA actually published for that entity. It is not a transliterator: a company with no registered English name/kana will not be found by a romanised query that NTA never recorded. prefecture is resolved against a static 47-prefecture table (src/prefectures.ts) so it works by Japanese name, English name, or code regardless of whether the matched company registered an English address — filtering on the register's own per-company prefecture_en column directly would silently miss almost every entity in a prefecture whose companies have no English name at all.

FTS5's trigram tokenizer windows by character, not byte: a query under 3 characters produces no trigrams and matches nothing via MATCH even when it is a real substring (verified: a 2-character Japanese query returns 0 rows via FTS, 5+ via LIKE, for an identical string). jp_company_search falls back to a plain substring scan for any query under 3 characters.

Refresh and dupe-check notes: see scripts/japan-corporate-number/README.md (operator-only, does not publish) and gbizinfo above.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "japan-corporate-number": {
      "url": "https://gateway.pipeworx.io/japan-corporate-number/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/japan-corporate-number/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Both URLs reach the same gateway and the same 1769+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/jp_corporate_number \
  -H 'Content-Type: application/json' \
  -d '{"number":"1000012160153"}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/jp_corporate_number. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Standalone (no gateway account)

This package also runs as a local stdio MCP server — no Pipeworx account, no gateway round-trip:

{
  "mcpServers": {
    "japan-corporate-number": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-japan-corporate-number"]
    }
  }
}

Or run it directly to confirm it starts:

npx -y @pipeworx/mcp-japan-corporate-number

It speaks MCP over stdin/stdout and answers initialize/tools/list/tools/call for only this pack's tools — none of the shared meta-tools the gateway connection above adds. Same source, same tools, no ask_pipeworx routing.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about Japan Corporate Number data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides
• pipeworx.io

License

MIT