edinet-tools

Python library for Japan's EDINET disclosure system — the official source for securities reports, shareholding notices, tender offers, and other regulatory filings from listed Japanese companies.

J-GAAP, IFRS, and US-GAAP filers tag the same figure under different XBRL elements. edinet-tools maps them all to one typed Python field.

Zero runtime dependencies. Typed parsers for all 42 EDINET document types.

import edinet_tools

toyota = edinet_tools.entity("7203")
docs = toyota.documents(days=30)   # requires EDINET_API_KEY (see Configuration)
report = docs[0].parse()  # → SecuritiesReport, LargeHoldingReport, etc.

Install

pip install edinet-tools

Requires Python 3.10+. Standard library only.

0.8.1 or later is required. EDINET moved its API to api.edinet-fsa.go.jp at the end of August 2026; earlier versions call the old host and cannot fetch anything.

Upgrading from 0.7.x? 0.8.0 contains breaking changes; see MIGRATING.md.

Design

edinet-tools has three layers:

1. API client — fetch document listings and download filings in any format (XBRL, PDF, HTML)
2. Typed parsers — every EDINET document type routes to a named Python dataclass with structured fields
3. Full capture — elements not yet mapped to typed fields are preserved in raw_fields, unmapped_fields, text_blocks, and raw_facts (the full XBRL fact set), so you can explore what's available and nothing is silently dropped

Each parser maps known XBRL elements to typed Python fields (dates, decimals, strings). As EDINET evolves or new elements become useful, adding a field is one line in the element map and one line on the dataclass.

If a filing doesn't state a figure, the field is None: never a guess, never a number borrowed from another accounting standard or from the parent company. Every mapping is tested against real filings and cross-checked against issuers' own earnings releases.

EDINET Document Types

EDINET defines 42 document types spanning corporate disclosure, capital markets activity, and governance reporting. edinet-tools provides typed parsers for all of them.

Amendments (even-numbered codes like 130, 150, 190) route to the same parser as their base type and set is_amendment = True.

from edinet_tools import supported_doc_types, doc_type

supported_doc_types()  # All 42 codes with typed parsers

dt = doc_type("235")
print(dt.name_en)  # "Internal Control Report"
print(dt.name_jp)  # "内部統制報告書"

Usage

Entity Lookup

import edinet_tools

toyota = edinet_tools.entity("7203")      # By ticker (digit or alphanumeric)
toyota = edinet_tools.entity("Toyota")    # By name search
toyota = edinet_tools.entity("E02144")    # By EDINET code
print(toyota.name, toyota.edinet_code)    # TOYOTA MOTOR CORPORATION E02144

# Look up by Japan Corporate Number (法人番号)
toyota = edinet_tools.entity_by_corporate_number("1180301018771")

# Name search handles full-width/half-width, gaiji (㈱), and middle-dot variants
mufg = edinet_tools.search("三菱UFJ銀行")  # matches the catalog's ＵＦＪ form too

banks = edinet_tools.search("bank", limit=5)

Entity data comes from FSA registry snapshots bundled with the package, so lookup and search work offline. Snapshots are refreshed each release. Loading one older than a year raises StaleDataWarning, and EntityClassifier accepts paths to newer CSVs if you download your own.

Fetching Documents

# All filings for a date (requires EDINET_API_KEY)
docs = edinet_tools.documents("2026-01-20")

# Filter by company and type
earnings = toyota.documents(doc_type="120", days=365)

Parsing

report = doc.parse()

# Securities Report — consolidated financials (J-GAAP, IFRS, US-GAAP)
report.net_sales
report.operating_cash_flow
report.roe
report.accounting_standard  # "Japan GAAP", "IFRS", or "US GAAP"
report.segments             # list[SegmentRow] — per-segment metrics

# Figures that depend on ownership basis come as explicit pairs (0.8.0)
report.net_income_owners    # attributable to owners of parent
report.net_income_total     # includes non-controlling interests
report.net_assets_total
report.net_assets_owners    # None for J-GAAP filers (never filed as one element)

# Large Shareholding Report
report.filer_name
report.target_company
report.ownership_pct        # joint filing → the co-filers' GROUP total, not
report.is_joint_filing      #   the named filer's own stake (~half are joint)
report.joint_holders

# Tender Offer
report.acquirer_name
report.target_name
report.holding_ratio_after

# Any report
report.fields()     # List available typed fields
report.to_dict()    # Export as dictionary
report.raw_fields        # All XBRL elements by element ID
report.text_blocks       # Narrative text block content
report.extraction_flags  # parse-time structural checks (0.8.0): impossible
                         # values are withheld as None, never served

Download Formats

from edinet_tools.api import fetch_document

csv_zip = fetch_document("S100ABC")            # XBRL CSV (default)
pdf = fetch_document("S100ABC", type=2)        # PDF
html_zip = fetch_document("S100ABC", type=1)   # HTML documents

Configuration

Get a free API key from EDINET (video walkthrough):

export EDINET_API_KEY=your_key_here

The key is read from the environment only. Entity lookup and parsing work without an API key (document fetching requires one).

Testing

pytest tests/ -v  # 1,000+ tests

Links

• Changelog
• PyPI
• GitHub
• EDINET

License

MIT

Independent project. Not affiliated with Japan's Financial Services Agency. Verify data independently before making financial decisions.