Syndicator Blog

RSS

SEC XBRL RSS Feeds Turn Every EDGAR Financial Filing Into a Subscribable Stream

Most people think of RSS as a blog thing. One of the busiest feeds on the web belongs to a regulator. Every time a US public company files financial statements with the SEC, the filing shows up in an RSS feed within minutes, with the company, the form type, the reporting period and links to the machine-readable data attached.

That data is XBRL, short for eXtensible Business Reporting Language. It's an XML format where every number in a financial statement carries a tag saying what it is: us-gaap:Revenues, us-gaap:AccountsPayableCurrent, and so on, plus the period and the unit. A 10-Q stops being a PDF you read and becomes a set of facts a program can query. The RSS feed is how you find out a new set of facts exists.

The four live feeds

The SEC lists its feeds on its structured disclosure RSS page. There are four:

FeedWhat it carries
https://www.sec.gov/Archives/edgar/usgaap.rss.xmlFilings with financial statements tagged in the US GAAP or IFRS taxonomies
https://www.sec.gov/Archives/edgar/xbrl-inline.rss.xmlThe same kind of filings, in Inline XBRL format
https://www.sec.gov/Archives/edgar/xbrl-rr.rss.xmlMutual fund risk/return summaries
https://www.sec.gov/Archives/edgar/xbrlrss.all.xmlEvery XBRL filing submitted to the SEC

They refresh every ten minutes, Monday to Friday, 6am to 10pm Eastern. The channel header of the US GAAP feed says it holds up to 200 items. Polling faster than every ten minutes gets you nothing, and outside those hours the feed doesn't move at all.

The SEC also warns that the content and format of these files may change without notice. Take that seriously and write a parser that tolerates missing elements.

What one item looks like

Each item is plain RSS 2.0 on the outside. The title is the company name, the CIK and the filer role. The link goes to the filing index page on EDGAR. The guid and the enclosure both point to a zip of the filing's XBRL files, so the feed uses the same enclosure mechanism podcasts use, only the payload is a financial statement instead of an MP3. The plain description holds the form type.

The useful part sits in an edgar:xbrlFiling block inside each item. It carries the company name, form type, filing date, CIK, accession number, file number, acceptance time, reporting period, the SEC office that reviews the company, its SIC code and its fiscal year end. Some items add otherCikNumbers for co-registrants. Then comes edgar:xbrlFiles, one edgar:xbrlFile per document, each with a sequence number, file name, type, size, description and URL as attributes. That list covers the main filing, exhibits, images and the schema and linkbase files.

Two quirks are worth knowing before you write code. The feed is encoded as windows-1252, not UTF-8. And in the monthly archive files the edgar namespace is declared on each xbrlFiling element, not on the root. Match elements by local name and both problems go away.

A watcher in 20 lines

This script reads the US GAAP feed and prints every 10-K and 10-Q with its period and filing link. It uses only the Python standard library, and the {*} wildcard (Python 3.8 and later) ignores whatever namespace URI the feed happens to use.

import sys
import urllib.request
import xml.etree.ElementTree as ET

FEED = "https://www.sec.gov/Archives/edgar/usgaap.rss.xml"
WANT = {"10-K", "10-Q"}

def load(src):
    if src.startswith("http"):
        req = urllib.request.Request(src, headers={"User-Agent": "Jane Doe [email protected]"})
        with urllib.request.urlopen(req, timeout=30) as r:
            return ET.fromstring(r.read())
    return ET.parse(src).getroot()

root = load(sys.argv[1] if len(sys.argv) > 1 else FEED)
for item in root.iter("item"):
    f = item.find("{*}xbrlFiling")
    if f is None or f.findtext("{*}formType") not in WANT:
        continue
    print(f.findtext("{*}formType"),
          f.findtext("{*}cikNumber"),
          f.findtext("{*}period"),
          f.findtext("{*}companyName"),
          item.findtext("link"), sep="\t")

We tested it against a windows-1252 file built to the feed's structure, with non-ASCII company names and items that skip the filter. Pass a local file path as the first argument to do the same.

Note the User-Agent. The SEC's EDGAR access rules ask automated clients to identify themselves with a name and contact email, and cap traffic at 10 requests a second. Scripts that send a default library user agent tend to get blocked. A feed you poll every ten minutes is nowhere near the rate limit, but the header still matters.

Don't lose filings when the window rolls

A 200-item feed is a sliding window. On deadline days, when thousands of companies file 10-Qs in the same week, that window turns over fast. A watcher that misses a few runs can miss filings for good.

The SEC learned this the hard way. Its early XBRL feeds held every filing ever submitted. In December 2008 they were cut to the most recent 100 items, and the SEC's own notice at the time admitted that users who didn't know about the change lost continuity. It shipped a one-off catch-up feed to fill the gap.

Today the backfill is the monthly archive. One file per month, named xbrlrss-YYYY-MM.xml, going back to April 2005 and covering all structured data filings across programs. Same item format, so the script above reads it unchanged. Store the accession numbers you've seen, and if your watcher was down, replay the month's archive and skip what you already have. Accession numbers are unique per filing, which makes them a better dedupe key than the guid.

The monthly files are big. The October 2026 file is over 3 MB of XML, so fetch it once and cache it rather than pulling it on every run. If you poll the live feeds, send conditional GET headers and let the server answer with a 304 when nothing changed.

From the feed to the numbers

The feed tells you a filing exists. To get the numbers without unzipping XBRL yourself, the SEC runs free JSON APIs on data.sec.gov, documented on its EDGAR API page. No API key. The CIK from the feed item, padded to ten digits, is all you need:

curl -s -H "User-Agent: Jane Doe [email protected]" \
  https://data.sec.gov/api/xbrl/companyconcept/CIK0000320193/us-gaap/AccountsPayableCurrent.json

That returns every value Apple has reported for that one tag, each with the period end date, the value, the accession number, the fiscal year and quarter, the form and the filing date. companyfacts returns all tags for a company in one call, and frames returns one tag across every company for a single period. The SEC says the APIs usually pick up a new filing in under a minute.

Two limits. The APIs only cover standard taxonomies like us-gaap, ifrs-full and dei, so company-specific custom tags are missing. And data.sec.gov doesn't send CORS headers, so you can't call it from browser JavaScript on your own site. Run it server-side or in a scheduled job.

Why this matters for syndication

The EDGAR feeds are a good model of what RSS does well when the publisher is serious about it. Stable URLs, a documented update schedule, typed extension data in its own namespace, an enclosure for the heavy payload and a permanent archive for anyone who falls behind. Plenty of blogs could learn from a federal agency here. The one thing it gets wrong is the sliding window with no paging, and the monthly archive is the patch for that.