Menu
Iceland Tax Authority — annual reports (ársreikningar), company registry, ownership chain mapping by kennitala.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Iceland energy authority — electricity generation, use, fuel sales, power plants and licences. Use for energy-system analysis.
Fiskistofa — public WFS layers for fishing closures, regulations and fishing areas; paid REST catch/quota data is excluded.
Hafrannsóknastofnun / MFRI — annual fish-stock assessments, advice, landings and survey series in embedded tables.
Environment Agency of Iceland GIS — open WFS layers for contaminated land, water, protected areas, noise and wastewater.
Icelandic Met Office (Veðurstofa) — weather observations, stations, forecasts and earthquakes via the modern JSON API at api.vedur.is.
Icelandic state accounts actuals (ríkisreikningur, Fjársýsla) — yearly revenue and expense 2015+, 36 málefnasvið policy areas.
| name | skatturinn |
| description | Iceland Tax Authority — annual reports (ársreikningar), company registry, ownership chain mapping by kennitala. |
Company registry and annual reports (ársreikningar) from the Icelandic tax authority.
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.
| 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) |
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.
Each company page at /fyrirtaekjaskra/leit/kennitala/{kt} shows:
Reports contain:
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.
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.
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.
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"}
Navigate to cart URL from response
Click "Áfram" button to proceed
Download triggers - clicking download button returns PDF
# 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
Observed behavior:
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.
# Test extraction from any PDF
uv run python scripts/skatturinn.py extract /path/to/report.pdf
The script looks for:
\d{6}-?\d{4}\d+%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.",
"ownership_pct": 100.0,
"type": "company",
"owners": [...] # Recursive
},
{
"kennitala": "0101801234",
"name": "Jón Jónsson",
"ownership_pct": 50.0,
"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 = owner_kt[0] in "4567"
owner_data = {
"kennitala": owner_kt,
"name": owner["name"],
"ownership_pct": owner["ownership_pct"],
"type": "company" if is_company else "person"
}
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
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:
PDF variability: Report formats vary by year and preparer (accountant). Some are structured, some are scanned images.
Ownership on page vs PDF: The company lookup page shows current beneficial owners. The annual report PDF shows ownership as of fiscal year end - these may differ.
Holding structures: Beneficial owners may be foreign entities without Icelandic kennitala, making chain tracing incomplete.
Rate limits: Unknown but assume conservative limits. Add 2-5 second delays between requests.
Terms of service: No explicit prohibition on automated access, but bulk scraping may trigger blocks. Consider contacting fyrirtaekjaskra@skatturinn.is for data access agreements.
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.
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
SQL queries in /evidence-reports/sources/skatturinn/:
-- ownership_network.sql
SELECT
parent_kt,
child_kt,
ownership_pct,
depth
FROM read_csv('../data/processed/skatturinn/ownership.csv')
# Download single report
uv run python scripts/skatturinn.py download --kennitala 5012043070 --year 2023
# Map ownership chain
uv run python scripts/skatturinn.py chain --kennitala 5012043070 --depth 3
# Bulk download for list of companies
uv run python scripts/skatturinn.py bulk --input companies.txt --year 2023