CAP Alert Extractor

Point at ANY CAP alerting authority and get structured emergency-alert rows.

This Actor is a generic runner over the OASIS Common Alerting Protocol (CAP) 1.2 — the international standard that national meteorological and emergency-management agencies use to publish public warnings (severe weather, floods, earthquakes, tsunamis, air-quality, civil emergencies). The WMO Register of Alerting Authorities lists ~130 authorities that publish CAP. Give this Actor any of their feeds — or a single CAP document — and get flat, analysis-ready rows. No per-country scraper needed.

Why this is one Actor, not a hundred

Every authority speaks CAP, but they serialize it three different ways on the wire. This Actor auto-detects and decodes all three, so a single input works everywhere:

Serialization	Example authority	What it looks like
GeoJSON	US NWS (api.weather.gov/alerts/active)	one features[] per alert, CAP fields in properties
Inline CAP-XML feed	UN GDACS (gdacs_cap.xml)	an RSS feed where each <item> embeds a full <cap:alert>
Atom/RSS → linked CAP docs	Canada NAAD, PH PAGASA, NZ MetService	a feed whose entries link to standalone CAP XML documents (auto-followed)

It also handles the details the CAP spec leaves to the publisher:

• the cap: namespace prefix is optional and inconsistent — elements are matched by local-name, so <cap:alert>, <alert>, and default-namespaced documents all parse.
• <area> geometry can be a polygon, a circle, or a geocode (SAME / FIPS / UGC) — all three shapes are captured; alerts with only a geocode are fine.
• status: Test / Exercise and msgType: Cancel / Update are real CAP values — the status filter (default Actual) keeps test messages out of your data, and msgType is surfaced so a Cancel is never mistaken for a live warning.
• multi-language alerts (Canadian EN+FR, EU 24-language) emit one row per language.

Input

Field	Type	Default	Description
startUrl	string (required)	—	A CAP feed URL (GeoJSON / CAP-XML / Atom-RSS) or a single CAP XML document URL
mode	enum feed | alert	feed	feed lists current alerts from a feed; alert parses one CAP document
status	enum	Actual	Only emit alerts with this CAP status (empty = any). Keeps Test/Exercise out
severity	enum	(any)	Filter by CAP severity (Extreme / Severe / Moderate / Minor / Unknown)
urgency	enum	(any)	Filter by CAP urgency (Immediate / Expected / Future / Past / Unknown)
certainty	enum	(any)	Filter by CAP certainty (Observed / Likely / Possible / Unlikely / Unknown)
maxRecords	integer	1000	Stop after this many rows
userAgent	string	(built-in)	Override the User-Agent (some authorities request a descriptive UA)
timeoutSecs	integer	60	Per-request timeout

Example inputs

{ "startUrl": "https://api.weather.gov/alerts/active", "mode": "feed", "status": "Actual" }

{ "startUrl": "https://www.gdacs.org/xml/gdacs_cap.xml", "mode": "feed", "severity": "Severe" }

{ "startUrl": "https://rss.naad-adna.pelmorex.com/", "mode": "feed" }

{ "startUrl": "https://alerts.metservice.com/cap/rss", "mode": "feed" }

Output

The natural billable row is (alert × info-language × area) — one CAP <alert> with 2 language <info> blocks each covering 7 <area> blocks yields 14 rows. Every row is a flat, fully-nullable superset carrying alert-level fields (identifier, sender, sent, status, msgType, scope, references, incidents), info-level fields (language, categories, event, responseTypes, urgency, severity, certainty, effective, onset, expires, senderName, headline, description, instruction, eventCodes, parameters), and area-level fields (areaDesc, polygons, circles, geocodes, altitude, ceiling), plus provenance (_sourceUrl, _capUrl, _mode, _rowIndex, and the lossless _raw CAP XML when applicable).

{
  "identifier": "urn:oid:2.49.0.1.840...",
  "sender": "w-nws.webmaster@noaa.gov",
  "sent": "2026-07-09T12:26:00-04:00",
  "status": "Actual",
  "msgType": "Update",
  "language": "en-US",
  "event": "Flood Watch",
  "severity": "Severe",
  "urgency": "Future",
  "certainty": "Possible",
  "effective": "2026-07-09T12:26:00-04:00",
  "expires": "2026-07-10T00:00:00-04:00",
  "senderName": "NWS State College PA",
  "headline": "Flood Watch issued July 9 ...",
  "areaDesc": "Perry; Dauphin; Lebanon; Cumberland; Adams; York; Lancaster",
  "geocodes": [{ "valueName": "SAME", "value": "042099" }, { "valueName": "UGC", "value": "PAZ056" }]
}

Honest behavior

• A reachable feed with 0 in-window alerts is a real result (exit 0, zero rows) — not an error.
• If every source is unreachable / unparseable, the Actor fails fast rather than emitting a fabricated empty success, so you can trust a 0-row run.
• It never invents rows and never reports an error as an honest empty.

Pricing

Pay-per-event: $0.0005 per alert record (one alert × info × area row), with a volume discount across subscription tiers.

Notes

• Metadata only — the Actor reads the alert content the authority publishes; it does not download any linked binary attachments.
• Discover more authorities and their feed URLs in the WMO Register of Alerting Authorities.