Skatturinn (Iceland Tax Authority)

Company registry and annual reports (ársreikningar) from the Icelandic tax authority.

Overview

Skatturinn operates the Company Registry (Fyrirtækjaskrá) and Annual Reports Registry (Ársreikningaskrá). Annual reports are free of charge since January 1, 2021. Downloading goes through a server-side shopping cart, but no browser is required — the cart is cookie/viewstate state that httpx handles directly. See "Access Method" below.

Use case: Map company ownership chains by extracting beneficial owners from annual reports, then recursively looking up parent company reports.

URLs

Resource	URL Pattern
Company lookup	https://www.skatturinn.is/fyrirtaekjaskra/leit/kennitala/{kennitala}
Search by name	https://www.skatturinn.is/fyrirtaekjaskra/leit?nafn={company_name}
Search page	https://www.skatturinn.is/fyrirtaekjaskra/leit
Annual reports info	https://www.skatturinn.is/fyrirtaekjaskra/arsreikningaskra/
API Portal	https://api.skatturinn.is/ (requires registration, limited endpoints)

Finding a Company's Kennitala

To find a company's kennitala by name, use WebFetch on the search URL:

https://www.skatturinn.is/fyrirtaekjaskra/leit?nafn={company_name}

Example: ?nafn=dansport returns the company page with kennitala 4807032350.

This is more reliable than Google searches. The page shows matching companies with their kennitala prominently displayed.

Data Available

Company Lookup Page

Each company page at /fyrirtaekjaskra/leit/kennitala/{kt} shows:

• Basic info: Name, kennitala, address, registration date, legal form
• Management: Directors, beneficial owners with ownership percentages
• Beneficial owners (Raunverulegir eigendur): Name, birth year/month, % and holding type (Beint/Óbeint). Beneficial owners are by definition natural persons — the page shows no kennitala for them, only birth year/month. Listed companies (kauphallarfélög) render no owner block at all.
• Shareholders (Hluthafar): Section on some pages; may carry kennitalas
• Activity: ÍSAT codes (business classification)
• Annual reports table: List of submitted reports with:
  • Operating year (Rek. ár)
  • Submission date (Skiladagsetning)
  • Report number (Nr. ársreiknings)
  • Report type (Tegund ársreiknings)

Annual Report PDFs

Reports contain:

• Balance sheet, income statement, cash flow
• Beneficial ownership (Raunverulegir eigendur): Name, kennitala, ownership %
• Shareholders (Hluthafar): For smaller companies
• Notes on related party transactions

Access Method: plain HTTP (httpx) — no browser

No public API exists for downloading annual reports; the site drives a server-side shopping cart. But it does not need a browser. scripts/skatturinn.py imports no browser library at all — the whole flow, download included, is httpx plus ASP.NET viewstate handling:

GET  /fyrirtaekjaskra/leit/kennitala/{kt}   -> scrape company + report rows
GET  addToCart?itemid=…&typeid=…            -> cart id (kid)
GET  cart page                              -> extract __VIEWSTATE
POST cart page                              -> ZIP/PDF bytes

The cart is session state in cookies, not JavaScript. httpx.AsyncClient keeps the cookie jar across those calls, which is all it ever needed.

Setup

uv sync

That is the whole setup. Do not install Chromium for this skill — earlier versions of this file told you to, and it was wrong for long enough to mislead both a human and an agent into planning browser-only monitoring. tests/health/test_skatturinn.py is a plain HTTP probe for the same reason.

Page Structure (Verified)

Beneficial Owners - in .collapsebox elements:

<div class="collapsebox">
  <span><h4>Viktor Ólason</h4></span>
  <div class="tablewrap">
    <table class="annualTable">
      <thead>
        <tr><th>Fæðingarár/mán</th><th>Búsetuland</th><th>Ríkisfang</th><th>Eignarhlutur</th><th>Tegund eignahalds</th></tr>
      </thead>
      <tbody>
        <tr><td>1964-JÚNÍ</td><td>Ísland</td><td>Ísland</td><td>100%</td><td>Beint eignarhald</td></tr>
      </tbody>
    </table>
  </div>
</div>

Annual Reports - table with td[data-itemid]:

<td data-itemid="808877" data-typeid="1">Ársreikningur<a href="#" class="tocart">Kaupa</a></td>

Report Type IDs:

typeid	Type	Download
1	Ársreikningur (PDF)	Works
4	Rafrænn ársreikningur (XBRL?)	Different flow, may fail

Some years have typeid=4 instead of 1. The current script only handles typeid=1 reliably.

Download Flow (Verified)

1. Add to cart via API:

   GET /da/CartService/addToCart?itemid={itemid}&typeid={typeid}
   

   Returns: {"addCartItemResult": true, "shoppingCartUrl": "https://vefur.rsk.is/Vefverslun/Default.aspx?kid=XXXX"}

2. Navigate to cart URL from response

3. Click "Áfram" button to proceed

4. Download triggers - clicking download button returns PDF

CLI Usage

# Get company info with beneficial owners
uv run python scripts/skatturinn.py info 5012043070

# Download specific year
uv run python scripts/skatturinn.py download 5012043070 --year 2024

# Download all available reports
uv run python scripts/skatturinn.py download 5012043070

# Map ownership chain (recursive)
uv run python scripts/skatturinn.py chain 5012043070 --depth 3

Anti-Bot Measures

Observed behavior:

1. Session cookies required - JSESSIONID must persist across requests
2. Rate limiting - 3 second delay recommended between requests
3. Cross-domain flow - Cart redirects from skatturinn.is to vefur.rsk.is

PDF Extraction

Report Types

1. Hnappurinn (Micro-company) - Simplified 4-page format without ownership section. Only financial statements.
2. Full ársreikningur - Complete reports with ownership tables, auditor statements, detailed notes.

For ownership chain mapping, the web page scraping is more reliable since it shows current beneficial owners directly. PDF extraction is useful for historical ownership or additional shareholder detail in full reports.

Extract Command

# Test extraction from any PDF
uv run python scripts/skatturinn.py extract /path/to/report.pdf

The script looks for:

• Kennitala patterns: \d{6}-?\d{4}
• Ownership percentages: \d+%
• Section keywords: "eigend", "hluthaf", "eignarhald"

Ownership Chain Mapping

The main use case: follow ownership through multiple levels.

async def map_ownership_chain(
    root_kennitala: str,
    max_depth: int = 5,
    visited: set[str] | None = None
) -> dict:
    """
    Recursively map ownership chain starting from a company.

    Returns nested structure:
    {
        "kennitala": "5012043070",
        "name": "Example ehf.",
        "owners": [
            {
                "kennitala": "1234567890",
                "name": "Parent Company hf.",
                "birth_year_month": None,
                "ownership_pct": 100.0,
                "ownership_types": ["Beint eignarhald"],
                "type": "company",
                "owners": [...]  # Recursive
            },
            {
                "name": "Jón Jónsson",
                "birth_year_month": "1980-JANÚAR",
                "ownership_pct": 50.0,
                "ownership_types": ["Beint eignarhald"],
                "type": "person"  # No recursion for individuals
            }
        ]
    }
    """
    if visited is None:
        visited = set()

    if root_kennitala in visited or max_depth <= 0:
        return {"kennitala": root_kennitala, "circular_ref": True}

    visited.add(root_kennitala)

    # Get company info from skatturinn
    info = await get_company_info(root_kennitala)

    # Download latest annual report and extract ownership
    # (page info may be stale - PDF has authoritative ownership)

    result = {
        "kennitala": root_kennitala,
        "name": info["name"],
        "owners": []
    }

    for owner in info["beneficial_owners"]:
        owner_kt = owner["kennitala"]

        # Determine if owner is company or person
        # Companies: first 2 digits are month (01-12), not day
        # Actually: company kennitalas start with 4-7 in first digit
        is_company = bool(owner_kt and owner_kt[0] in "4567")

        owner_data = {
            "name": owner["name"],
            "birth_year_month": owner["birth_year_month"],
            "ownership_pct": owner["ownership_pct"],
            "ownership_types": owner["ownership_types"],
            "type": "company" if is_company else "person"
        }
        if owner_kt:
            owner_data["kennitala"] = owner_kt

        if is_company:
            # Recurse into parent company
            owner_data["owners"] = (await map_ownership_chain(
                owner_kt, max_depth - 1, visited
            )).get("owners", [])

        result["owners"].append(owner_data)

    return result

Kennitala Format

Icelandic identification numbers:

Type	Format	First digit	Example
Person	DDMMYY-NNNN	0-3 (day)	0101801234 (Jan 1, 1980)
Company	DDMMYY-NNNN	4-7	5012043070 (Dec 10, 2004)

The first 6 digits encode the registration date. Century determined by 9th digit:

• 8 = 1800s, 9 = 1900s, 0 = 2000s

Data Caveats

1. PDF variability: Report formats vary by year and preparer (accountant). Some are structured, some are scanned images.

2. Ownership on page vs PDF: The company lookup page shows current beneficial owners — always natural persons, with birth year/month but no kennitala (Owner.kennitala is None; the chain cannot recurse through them from the page alone). The annual report PDF shows ownership as of fiscal year end with full kennitalas — run chain with --download to get kennitalas and follow corporate structures. Page and PDF may differ.

3. Holding structures: Beneficial owners may be foreign entities without Icelandic kennitala, making chain tracing incomplete.

4. Rate limits: Unknown but assume conservative limits. Add 2-5 second delays between requests. Rapid consecutive requests bounce to a terms page (Notkunarskilmálar); get_company_info() detects that and returns None instead of misreporting it as a company — wait and retry. The name-search endpoint (?nafn=) can be flaky under load; the kennitala lookup page is the stable entry point.

5. Terms of service: No explicit prohibition on automated access, but bulk scraping may trigger blocks. Consider contacting fyrirtaekjaskra@skatturinn.is for data access agreements.

6. Session cookies: The shopping cart requires maintaining session state. Reuse one httpx.AsyncClient (and therefore one cookie jar) across addToCart → cart → download; a fresh client between those steps loses the cart.

7. Control without a percentage: Eignarhlutur can be - when a person qualifies as a beneficial owner through control rather than a stated equity percentage. In that case ownership_pct is null; inspect ownership_types for values such as Bein stjórnun or Tilnefning stjórnarmanna.

Evidence Integration

Store extracted data in /data/processed/skatturinn/:

/data/
  /raw/skatturinn/
    /{kennitala}_{year}.pdf     # Original PDFs
  /processed/skatturinn/
    /companies.csv              # Company metadata
    /ownership.csv              # Ownership relationships
    /ownership_chains.json      # Full chain mapping

Example query — ownership network (DuckDB):

SELECT
    parent_kt,
    child_kt,
    ownership_pct,
    depth
FROM read_csv('../data/processed/skatturinn/ownership.csv')

Quick Commands

# Download single report (kennitala is positional, not a --flag)
uv run python scripts/skatturinn.py download 5012043070 --year 2023

# Map ownership chain
uv run python scripts/skatturinn.py chain 5012043070 --depth 3

# Company info incl. beneficial owners (ownership check)
uv run python scripts/skatturinn.py info 5012043070

# Bulk download for list of companies (no bulk subcommand — loop instead)
while read kt; do uv run python scripts/skatturinn.py download "$kt" --year 2023; done < companies.txt

Related Skills

• hagstofan - Statistics Iceland for economic context
• sedlabanki - Central Bank for financial sector data
• National Registers API - Basic company info (no annual reports)