Official Japanese government data, structured for the English-speaking world.
Filings, financial statements, subsidies, real-estate prices, company records, laws — and the business-day calendar — read from official government sources only and returned as clean snake_case JSON, with every Japanese original preserved. No scraping. No broken selectors. Tested against frozen datasets on every release.
From 有価証券報告書 to snake_case.
Japan's public data is rich, official — and hard to use from outside: Japanese-only fields, wareki dates, full-width digits, amounts in thousands of yen. Each Actor does one job: turn an official record into JSON you can put straight into a pipeline.
{ "doc_id": "S100YIZC", "filer_name_ja": "山口放送株式会社", "filer_name_en": null, "doc_type": "Annual Securities Report", "period_start": "2025-04-01", "period_end": "2026-03-31", "submitted_at": "2026-06-30T09:00:00+09:00", "financials": { "net_sales": 4928920000, "net_income": 100436000, "number_of_employees": 120, "financials_basis": "non_consolidated" }, "attribution": "出典：金融庁 EDINET" }
Real output from the Japan Company Filings Actor. Note the null:
this filer registers no English name, so none is invented. Every generated or derived value in the
family carries confidence and
method — you always know whether a
field came from the API, a rule, or a model.
One Actor per source. Priced per record.
Each Actor covers one official system, with a flat input schema, a prefilled example that returns results in seconds, and a README that states what it does not do before what it does.
金融庁 有価証券報告書 $0.005 / filingenriched $0.079 弐 Japan Subsidies & Grants Data (Official) Which Japanese companies received which government subsidies — by corporate number, company name, or ministry. Amounts in JPY, ministry names in English. No LLM anywhere. METI · gBizINFO v2
経済産業省 補助金 $0.004 / record 参 Japan Real Estate Transaction Prices (MLIT Official) Quarterly actual and contract prices across all 47 prefectures, bilingual by source. Derived ㎡ unit prices, optional municipality × quarter aggregates at no extra charge. MLIT · Real Estate Library
国土交通省 取引価格 $0.003 / recordaggregates free 肆 Japan Company Data Enrichment (Official Registry) Corporate number lookup and firmographics: capital, employees, industry, establishment date, subsidy / procurement / patent counts. English names only when officially registered. gBizINFO + NTA Registry
法人番号・Gビズインフォ $0.006 / companyenriched $0.019 伍 Japan Laws & Regulations (e-Gov Official) Full text of any Japanese law as a structured article tree, at any point in time. Optional English reference translations — with an honest pointer to the official JLT database first. Digital Agency · e-Gov API v2
デジタル庁 法令 $0.004 / articletranslated $0.029 陸 Japan Financial Statements (EDINET Official) Normalized balance sheet, income statement and cash flow — Japan GAAP and IFRS — with the source XBRL element ID attached to every number, so each value is machine-verifiable. FSA · EDINET API v2
金融庁 財務三表 $0.03 / reportprior year included 漆 Japan Business Days & Calendar (Cabinet Office Official) Check, add and count Japanese business days — national holidays plus your own company closures — and convert wareki era dates both ways: 令和8年7月11日, R8.7.11 and reiwa 8 all parse. Deterministic, no LLM, no runtime API calls. Cabinet Office · holiday data
内閣府 国民の祝日 $0.003 / recorddeterministic · no LLM
doc_id values from Company Filings
straight into Financial Statements. Headline KPIs and the full three statements deliberately don't overlap —
you never pay twice for the same field.
Actors are published in phases — the Apify developer profile lists what is live today.
Three principles, engineered in — not written on.
Built not to break
No DOM selectors exist to break: every byte comes from a documented government API
or an official published dataset, fetched serially with per-source rate limits and
exponential backoff. One bad record never kills a run — failures are output as
explicit _error rows, unbilled.
Bilingual by design
English snake_case fields are canonical; the Japanese original always rides along in
*_ja. Wareki dates become ISO 8601, thousands-of-yen become raw JPY,
full-width digits are normalized — and every release is regression-tested against
frozen real API responses.
Honest about limits
Every README opens with a “What this does NOT do” section. Values a source
doesn't report are null, never estimated. Anything a model generated is
labeled method: "llm" with a confidence score — and verified against the
source text before it ships.
The same list we print at the top of every README — because a data product that oversells is a data product you can't build on.
- No scraping, ever. If an official API doesn't expose it, we don't have it.
- Not exhaustive. Coverage is exactly what each ministry publishes — no more.
- No guessed numbers. Unmapped, ambiguous, or bracketed values return null.
- No invented English names. Registry English or null; transliterations are opt-in and labeled.
- No bulk law corpus. Sequential full-corpus downloads are prohibited upstream; we point you to the official bulk XML instead.
- Not the only holiday list. Plain holiday lists have free alternatives; the calendar Actor is for business-day math, wareki parsing, and custom closures. Dates outside its covered range return explicit errors, never guesses.
- Not advice. Summaries and translations are reference material — not investment or legal advice.
Legible to agents, not just to people.
Every Actor is enabled for agentic users on Apify and shaped for tool calls: an agent can discover, evaluate, and run one — including payment — without a human in the loop. Deadline math and era conversion, the classic LLM stumbling blocks, are delegated to a deterministic tool instead of guessed.
- 入力最小Flat input schemas with the fewest possible required fields — prefills alone return meaningful results.
- 即応Default inputs answer in about 30 seconds, with hard per-run caps so a tool call can't run away.
- 構造出力Small, predictable JSON records — schema-versioned, with
element_mapandcoveragewhere numbers need verifying. - 価格明示Per-event prices are written in the README an agent reads before it runs — no surprises, and x402 autonomous payment is enabled.
Pay per event. Try before you pay.
Every Actor starts at $0.02 per run, then charges only for the records it actually delivers. Error rows are never billed.
| Actor | Per record | LLM tier (opt-in) | Free every run |
|---|---|---|---|
| Company Filings | $0.005 / filing | $0.079 enriched summary | 3 documents |
| Financial Statements | $0.03 / annual report | — no LLM by design | 3 documents |
| Subsidies & Grants | $0.004 / subsidy record | — no LLM by design | 50 records |
| Real Estate Prices | $0.003 / transaction | — aggregates free | 50 records |
| Company Enrichment | $0.006 / company | $0.019 enriched | 20 companies |
| Laws & Regulations | $0.004 / article | $0.029 translated | 20 articles |
| Business Days & Calendar | $0.003 / record | — no LLM by design | 50 records |
The free allowance is a real evaluation, not a teaser: enough records to judge the output quality, not enough to run a business on.
Charge limits are respected gracefully. If a run hits the maximum spend you set, it stops cleanly and keeps every record already fetched.