Medium Article Feed Scraper (Tag, Author, Publication)

Scrapes article metadata from any public Medium feed — by tag, by author, or by publication. No login, no API key, no cookies.

Useful for content research, competitor tracking, newsletter curation, trend monitoring on a topic, and building a reading index across many tags at once.

Three feed types

Type	Target looks like	Returns
tag	python, machine-learning	Recent articles carrying that tag, across all of Medium
user	dhh, @dhh	Recent articles by that author
publication	the publication slug from its URL	Recent articles in that publication

What this returns — and what it doesn't

This Actor returns article metadata plus the feed's own summary: title, author, tags, publication date, canonical URL, and a plain-text excerpt. It does not return full article bodies — every row carries a canonical url if you need to open the article itself.

What makes this different

CDATA is unwrapped, HTML is stripped. Medium's feed wraps nearly every value in <![CDATA[...]]> and renders excerpts as HTML. A scraper that passes those straight through puts literal <![CDATA[ markers in every title and <p> tags in every summary. Here titles come out as titles and summaries as plain text.

Both timestamps are actually comparable. Medium mixes two date formats inside a single feed item: published_at arrives as RFC-822 (Sun, 23 Aug 2026 03:11:27 GMT) while updated_at arrives as ISO-8601 with fractional seconds. Both are normalized to UTC ISO-8601, so you can sort and diff them without a second cleanup pass.

Clean URL and tracking URL, kept separate. Every feed link carries an RSS tracking suffix (?source=rss------python-5). url is the canonical link — stable, dedupe-able across feeds — and url_with_tracking preserves the original if you need it.

A stable article ID. Derived from the hex suffix of the article slug, so the same article surfaced by three different tag feeds dedupes to one ID rather than three rows.

excerpt_length tells you what you're looking at. The summary is capped at 500 characters, and this field reports the excerpt's true length — so a genuinely short teaser is distinguishable from a truncated one, instead of you having to guess.

Incremental mode. Remembers article IDs between runs, so a scheduled watch on a tag returns only what's new — and you are charged only for the new rows.

Input

Field	Type	Notes
feedType	enum	tag, user, or publication.
targets	array	Tags / usernames / publication slugs. Each fetched independently.
maxItems	integer	Per target. 0 = everything the feed returns.
incremental	boolean	Only articles not seen in previous runs.
proxyConfiguration	object	Not normally needed.

One limit worth knowing up front

A Medium feed is a fixed window of the most recent posts — there is no pagination parameter to page further back. maxItems can narrow that window but cannot extend it. To build history on a tag, run this on a schedule with incremental enabled and let the dataset accumulate.

Output

{
  "article_id": "medium.com:214bb62adf43",
  "title": "I Thought I Knew Python Until I Built My First Real Automation System",
  "author": "Muhummad Zaki",
  "published_at": "2026-08-23T04:27:13Z",
  "tags": ["technology", "programming", "coding", "python"],
  "summary": "The moment Python stopped being a programming language and became an employee...",
  "excerpt_length": 123,
  "url": "https://python.plainenglish.io/i-thought-i-knew-python-...-214bb62adf43",
  "feed_type": "tag",
  "feed_target": "python"
}

Each run also writes a RUN_COVERAGE record to the key-value store with what was requested, what came back, and any per-target failures — so a partial run is visible rather than silent.

Local development

pip install -r requirements.txt
python test_local.py python --type tag --out sample_output.json
python test_local.py dhh --type user --max 5

sample_output.json is real output from a live run of the python tag feed.