Home › Blog › X v11.90 Articles Scraper

X v11.90 Articles Scraper: threadgrab Long-Form 2026

June 11, 2026 · 8 min read · by ThreadGrab

If you scrape X Articles with a script that worked in March 2026, that script probably stopped working in late May. The X mobile app's v11.90 release moved the See All Articles entry from a buried profile-page tab to a top-level segmented control, and behind the scenes it swapped the public REST endpoint for a new GraphQL route. Scrapers that hard-coded the old URL now get 200 with an empty body, the article id format changed from 18-digit numeric to 19-digit base64, and the article body is now a JSON blob with named-entity spans instead of plain text. This post walks through what changed, why your old scraper broke, and how to use the threadgrab API to fetch the new v11.90 layout as clean Markdown in one call.

threadgrab has been updated to handle the v11.90 changes natively. Below is the new field reference, two ready-to-run code examples, and a side-by-side with the two other approaches developers have tried (raw X GraphQL, and the tweet.md project). If you maintain a public X-archive tool or build LLM training data from long-form creators, this is the shortest path to working output in 2026.

What X v11.90 Changed in the Articles Layout

Three things moved, and they are independent — a scraper can survive one or two of them but not all three at once.

1. The See All Articles entry moved

Before v11.90, the "See All Articles" link was a tab on the creator's profile page, three rows down below Posts, Replies, Media, and Likes. In v11.90 it became the first item in a new segmented control at the top of the profile header, sitting next to Posts / Replies / Media. The visible change is small, but the underlying request changed: the old tab called /Articles/search.json?include_replies=false, the new entry calls /i/api/graphql/<hash>/ArticleTimeline. Anything hard-coded to the old path now gets 200 with an empty array.

2. The article id format changed

The legacy article id was an 18-digit numeric string (for example 1823456789012345678). The new id is a 19-character base64 string with mixed case and digits (for example a1B2c3D4e5F6g7H8i9J0). The new id is also opaque — it does not encode the publication timestamp the way the old one did, so a scraper cannot infer the publish date from the id alone. URLs that worked in March now 404 because the id fragment is rejected by the routing layer.

3. The article body is now a named-entity JSON list

The old body was plain text with markdown-like asterisks for bold (**text**) and angle-bracket URL patterns for links. The new body is a JSON array of entity spans, where each span has a type field (text, bold, link, mention, hashtag, media_ref) and a payload. Naive text scrapers lose the formatting because they concatenate the text spans without traversing the bold, link, and mention spans. A bold phrase like **Freedom of Reach arrives as three separate spans: a text span containing the empty string, a bold span containing Freedom of Reach, and a trailing text span. Reassemble them wrong and the bold is gone.

How threadgrab Handles the New Layout

threadgrab's profile API wraps all three changes. You pass a public article URL or a username, and threadgrab returns a flat Markdown body plus metadata. The conversion logic walks the named-entity spans in document order and produces standard Markdown: text becomes plain text, bold becomes **text**, link becomes [text](url), mention becomes @username, hashtag becomes #tag, and media_ref becomes ![alt](image_url) with the alt text pulled from the span's alt_text field when present.

The result is a Markdown string that round-trips cleanly through any renderer (md2rich, GitHub, Obsidian, VS Code preview, md2pdf). threadgrab does not invent formatting the original author did not write — if the author did not bold a phrase, the output does not bold it.

Single article fetch (curl)

curl -s "https://threadgrab.com/api/profile?url=https://x.com/jack/status/a1B2c3D4e5F6g7H8i9J0" \
  | jq '.article | {title, author, published_at, body_markdown}'

The response is a JSON object with five top-level fields: title (string), author (object with username, display_name, verified), published_at (ISO 8601 string), body_markdown (the formatted Markdown), and media (array of {type, url, alt_text} objects for images, videos, and polls). Average response time on a US-east Cloudflare edge: 1.1 to 1.4 seconds for a 2,000-word article, 0.6 to 0.9 seconds for a 500-word article.

Batch profile scrape (curl with cursor pagination)

USERNAME="naval"
CURSOR=""
LIMIT=20

while true; do
  RESP=$(curl -s "https://threadgrab.com/api/profile/${USERNAME}/articles?cursor=${CURSOR}&limit=${LIMIT}")
  echo "$RESP" | jq '.articles[] | {title, published_at, id: .article_id}'
  CURSOR=$(echo "$RESP" | jq -r '.next_cursor')
  [ -z "$CURSOR" ] || [ "$CURSOR" = "null" ] && break
  sleep 1
done

This loop walks the creator's full Articles feed page by page. Each call returns 20 articles plus a next_cursor string; when next_cursor is empty or null, you have hit the end of the public feed. The one-second sleep between calls keeps the request rate at the 1 req/s/IP soft limit. A creator with 200 articles takes about 11 seconds end-to-end, including the sleep overhead.

Field Reference: threadgrab v11.90 Response Shape

The full JSON object returned by the single-article endpoint has the structure below. The fields are stable as of threadgrab v1.18 (June 2026).

The body_markdown field is the one most callers care about. It is a single string with newlines preserved from the source, and the bold / link / mention / hashtag / image syntax is rendered as standard Markdown that any renderer will interpret correctly. If you need plain text only, pass the result through sed or your favorite Markdown-to-text tool to strip the syntax — the original text content is recoverable in one pass.

Comparison: threadgrab vs Raw GraphQL vs tweet.md

Three approaches are in use today for scraping X Articles after the v11.90 change. Here is how they compare on the three things that broke.

For a use case that is "I want to read this Article offline as Markdown" or "I want to build a corpus of long-form posts for LLM training," threadgrab is the shortest path. For a use case that is "I am building a real-time X analytics product and need raw engagement data," the GraphQL route is unavoidable. tweet.md is a good fit for short posts and threads, less so for the new Article body.

Limitations and What threadgrab Does Not Do

Being honest about the edges of the tool matters more than overselling. The list below is what threadgrab v1.18 does not do, as of June 2026.

• Protected accounts. If the author's Articles are behind a login wall, threadgrab returns a 403. There is no workaround — the content is not on the public web.
• Edit history. An Article that the author has edited five times is returned in its current form. Earlier versions are not exposed by X, so threadgrab cannot reconstruct them.
• Quote-Article composite posts. A Quote-Article is a normal post with an Article card attached. threadgrab returns the post text and the card metadata, but not the embedded Article's body unless you make a second call to the article endpoint with the card's article_id.
• Drafts and scheduled Articles. Drafts and scheduled posts are not on the public web, so they are not scrapeable through any method, including threadgrab.
• Aggregate reach and per-audience numbers. The reach black box we covered in the Freedom of Reach post applies to threadgrab as well — the engagement snapshot in the response is what X shows to the public, not the internal reach number.

FAQ

Try It on a Public Article

The fastest way to see the new v11.90 layout round-trip through threadgrab is to pick any public Article on x.com, copy the URL, and run the single-article curl example above. You should get a clean JSON response within 1.5 seconds on a US or EU edge, with the body in Markdown that you can paste into Obsidian, a GitHub gist, or the md2rich converter for a side-by-side preview of how the formatting renders.

Related Reading

• X to Markdown 2026: 3 Ways to Save X Threads — covers the tweet.md, threadgrab, and browser-reader approaches for short posts and threads.
• 2026 X Thread Archiving: 5 Tools Compared — a wider look at Nitter, Thread Reader, archive.today, curl scripts, and threadgrab for full thread archival.
• X Articles Freedom of Reach 2026: The Algorithm Black Box — what the reach number does and does not tell you about who saw an Article.
• X Articles vs Bluesky vs LinkedIn Newsletter — the 2026 long-form showdown, with a decision tree for which platform to write on.