Threads Scraper: Posts, Search, Tags, Replies

Threads Scraper: Posts, Search, Tags, Replies

Scrape Threads (threads.com, Meta's text app) without a login: the posts of any profile, keyword and tag search results, the replies under a post and the profile itself. One row per post with the text, the author, the time, likes, replies, reposts, quotes, media and links. Pay per post, from $0.99 per 1,000.

Built for social listening and brand monitoring, creator and competitor tracking, research and AI data sets, and scheduled monitors: name a memory and every later run delivers and charges only the posts it has not delivered before.

Quick start

1. Open the Actor, type one or more Threads usernames (zuck, @mosseri, or a profile URL) into Profiles, or search words into Search words, or post URLs into Posts.
2. Set the caps: posts per profile (default 50), posts per search (default 100), replies per post (default 100), posts per run (default 1,000).
3. Run. Rows land in the dataset as JSON, CSV, Excel or through the API; the free coverage row per input says what was read and why it stopped.
4. For a monitor: give a Memory name and schedule the run. The profile row (followers, bio) comes on every run; posts come once.

What you get

One post row per Threads post, whichever way it was found (a profile feed, a keyword or tag search, a reply under a post, a post URL). Fill rates measured on 1,393 distinct search posts and 226 profile posts read on 2026-10-07:

Field	What it is	Filled
postId, code, url	the numeric id, the short code and the post's page	100%
text	the full text of the post	99.7% (media only posts have none)
postedAt, postedAtUnix	when it was published (UTC)	100%
authorUsername, authorId, authorFullName, authorVerified, authorPrivate, authorProfilePic, authorUrl	the author	100%
likeCount, replyCount, repostCount, quoteCount, reshareCount	engagement at the time of reading	100% (reshareCount is often null)
viewCount	the view count the public post page shows; on posts looked up by URL only	100% of posts by URL, null elsewhere (Threads shows no view count in feeds or search without a login)
language, languageTag, languageSource, languageUnknown	the language as an ISO code: Threads' own label when it has one (Chinese, Japanese and Korean posts), else a statistical classification of the text when it is 80 characters or more (languageSource: "text"); unknown when neither gives one. Measured: on 473 English posts of two profiles 370 were classified, 0 wrong, 103 left unknown (short texts); on 120 search and reply posts classified from their text (many languages, odd scripts) 9 labels were wrong, about 1 in 13, every English label right	78% of two profiles' posts carry a language; 55% on the "nike" search page (short posts)
mediaType, hasMedia, imageUrl, images, videos, carousel	text, image, video or carousel, with the media URLs (they expire after some days)	59% of search posts carry an image, 25% a carousel, 13% a video
linkAttachment	the link card (url, title, image)	1.7% of search posts have one
isReply, replyToUsername, replyToUserId, rootPostAuthorUsername, parentPostCode	reply context	23% of search posts are replies
isQuote, quotedPost, repostedPost	the quoted or reposted post (code, url, author, text, time)	when present
isPinned, isPaidPartnership, likeAndViewCountsDisabled, hasAudio, accessibilityCaption	flags	100% (isPaidPartnership was null on every logged out row)
source, sourceUsername, sourceQuery, sourceWindow	which input produced the row: profile, search, tag, replies or post	100%

One profile row per username looked up: userId, username, url, fullName, verified, private, followerCount, biography, bioLinks, profileTags, profilePic, profilePicHd, threadsBadge, recentPostCount (5 of 5 profiles read in the spike carried id, name, followers, bio, verified and the private flag; bio links were set on 1 of 5).

Free rows: one coverage row per profile, search, tag or post (pages read, posts read, repeated, known to the memory, filtered and why, the stop reason, for a post the reply count against the replies read, for a search the windows read), a status row for a username or post the site does not have, a private profile, a failed lookup, an unusable value or inputs the run did not start, and one summary row per run with every counter and what was charged.

How deep it goes

• Profiles: the profile page, then Threads' own feed with its cursor, newest first, to the cap, the start date or the end of the profile (a 226 post profile back to 2023 was read in 15 calls). A private profile gives its profile row and a status row; its posts are not public.
• Search: the logged out search answers one page of top posts per query (about 20 to 35) and ignores every paging request. Depth comes from searchMode: the recent page (about 12 newest posts, no overlap with top), or one top page per day of the date range (8 to 20 posts per day for a common word; 14 days of "nike" gave 179 distinct posts). The window end is exact; the start is loose (a window can carry a few older posts, which the date filter removes). Some terms answer "No results" when logged out. No completeness claim is made for a search: the coverage row says what was read.
• Replies: the reply threads the public post page shows (about 10 threads with their nested replies, 7 to 18 replies); Threads asks for a login to show more, which this Actor never uses. The coverage row carries the post's reply count against the replies read.

Every filter (dates, languages, minimum likes, replies, media) is applied to the final row in code, never trusted to the server, and a filtered post is never charged.

How much does it cost to scrape Threads?

Event	FREE	BRONZE	SILVER	GOLD
Post (per 1,000 post rows)	$1.10	$1.05	$1.00	$0.99
Profile (per 1,000 profile rows, one per username looked up and found, per run)	$2.50	$2.50	$2.50	$2.50
Search page (per 1,000 search pages answered with at least one post: the top page, the recent page or one day window)	$2.00	$2.00	$2.00	$2.00
Actor start (per run, per GB of memory; 256 MB runs pay one)	$0.005	$0.005	$0.005	$0.005

Platinum and Diamond plans pay the Gold price. The Store pricing card shows these same prices per 1,000 events: "$2.50 / 1,000" on the Profile row means one profile row costs a quarter of a cent, and "$2.00 / 1,000" on the Search page row means one search page costs a fifth of a cent.

Worked examples at the Gold price: one profile with its newest 50 posts costs $0.005 (start) + $0.0025 (profile) + $0.0495 (50 posts) = $0.057. A keyword search of one top page with 30 posts costs $0.005 + $0.002 + $0.0297 = $0.037. A daily monitor of 10 profiles under a memory name that finds 5 new posts costs $0.005 + $0.025 + $0.005 = $0.035.

What is free: repeated posts, posts the memory already delivered, filtered posts, search pages with no post, a username or post the site does not have, a private profile's status row, refused calls, coverage rows, the summary, everything after a platform restart, and every row of a run whose named memory could not be used.

Unpaid work is bounded: a run may spend 40 seconds plus 6 seconds per charged post and 8 seconds per charged profile or search page on work that charges nothing (usernames or posts the site does not have, searches with no result, pages whose posts are all filtered or already known). When that time is spent the inputs not started are listed in one free skipped row and the run ends. For a scheduled monitor keep the profile row on: it pays for the profile's page even when no post is new. With the profile row off and a date filter that nothing passes, a long list of quiet profiles is cut by this rule.

Input

Field	Meaning
usernames	Threads handles or profile URLs (up to 200). Each gives a profile row and its newest posts up to maxPostsPerProfile (0 reads the profile row only). A post URL pasted here is read as a post.
searchQueries	search words (up to 100), each read per searchMode; maxPostsPerQuery caps the posts kept per query.
tags	Threads tags, with or without the # (up to 100), searched on the tag surface.
postUrls	post URLs or post codes (up to 200): the post itself with its view count, then the reply threads the public page shows, up to maxRepliesPerPost (0 reads the post only).
searchMode	top (default, one page), recent, top_and_recent, daily_windows (one top page per day between postedAfter and postedBefore; the last 7 days when no dates are given).
postedAfter, postedBefore	UTC days (YYYY-MM-DD, today, yesterday, "7 days", "2 weeks", "1 month"); applied to every row; a profile walk stops once it reaches older posts.
languages	ISO codes (en, es, ja); posts with an unknown language are kept and marked unless dropUnknownLanguage is on.
minLikes, excludeReplies, onlyWithMedia	engagement and content filters.
includeProfile	the profile row per username (on by default).
maxPosts	the cap of post rows for the whole run (default 1,000).
memoryName, resetMemory	pay once per post across runs: the memory is a named key-value store this Actor creates in your account (one per listing); resetMemory forgets it before the run.
proxyConfiguration	Apify datacenter proxy by default, one exit address per run. Residential is not needed.

A value is used as sent or refused: an unusable date, a cap that is not a whole number, a switch that is not true or false, an unknown search mode or an empty list ends as one free invalid status row and no paid work.

API examples

curl -X POST "https://api.apify.com/v2/acts/automation_craft~threads-scraper/runs?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"usernames": ["zuck"], "maxPostsPerProfile": 50}'

const { ApifyClient } = require('apify-client');
const client = new ApifyClient({ token: 'YOUR_TOKEN' });
const run = await client.actor('automation_craft/threads-scraper').call({ searchQueries: ['sourdough starter'], searchMode: 'daily_windows', postedAfter: '7 days', languages: ['en'] });
const { items } = await client.dataset(run.defaultDatasetId).listItems();

from apify_client import ApifyClient
client = ApifyClient('YOUR_TOKEN')
run = client.actor('automation_craft/threads-scraper').call(run_input={'postUrls': ['https://www.threads.com/@zuck/post/Dd1MqfcG0aH']})
items = client.dataset(run['defaultDatasetId']).list_items().items

What this Actor does NOT do

• It does not log in and uses no cookies of its own or yours, so it reads what the public web shows: no followers or following lists, no private profiles' posts, no replies beyond the first reply threads of a post page.
• It does not deliver view counts on feed and search rows (Threads shows none to the logged out web); only a post looked up by URL carries the view count its page shows.
• It does not walk a keyword search past the one page the logged out search answers; depth comes from the recent sort and from day windows, and no search is reported as complete.
• It does not read a user's replies tab or reposts tab, and does not search for profiles by keyword.
• It does not label short texts: Threads' own label is used when present, texts of 80 characters or more are classified statistically (0 wrong of 370 English profile posts measured, about 1 in 13 wrong on a mixed language search sample; the source is on every row), and the rest is marked unknown and kept unless you ask to drop it.
• It does not use residential proxies by default and never calls a third party data service.

FAQ

How do I scrape Threads posts from a profile?

Put the username or the profile URL into usernames and set maxPostsPerProfile. The Actor reads the profile page, then Threads' own feed with its cursor, newest first, until the cap, the postedAfter date or the end of the profile. Each post is one row; the profile row carries the follower count, the bio and the links.

Can I search Threads by keyword or hashtag?

Yes. searchQueries reads the keyword search, tags the tag surface. The logged out search answers one page of top posts per query; searchMode: "daily_windows" reads one top page per day of your date range and recent the newest posts. The coverage row per query lists every page read and how many posts it held.

Does it scrape replies and comments on a Threads post?

Yes, the reply threads the public post page shows (about 10 threads with their nested replies), each as a post row with isReply: true and parentPostCode. Threads asks for a login to show more replies, which this Actor does not use; the coverage row says how many replies the post has against how many were read.

Why do some rows have no language?

Threads labels only Chinese, Japanese and Korean posts. Other posts are classified from their text when it is at least 80 characters long (languageSource: "text", a statistical classification: 0 wrong of 370 English profile posts, about 1 in 13 wrong on a mixed language search sample; shorter texts were wrong more often, which is why they are not classified); shorter texts are languageUnknown: true and kept, unless dropUnknownLanguage is on.

How does the monitor memory work?

Give a memoryName. The Actor creates a named key-value store in your account and remembers every post it delivered under that name; later runs with the same name deliver and charge only new posts, and still deliver the profile row with the current follower count. A scheduled run reads the newest posts and stops at the first page it already delivered. When a run is cut by maxPostsPerProfile, maxPosts or your charge limit before it reached the end, it remembers where it stopped, and the next run with the same name continues from there after the newest posts, so a burst of new posts is never lost, only spread over the following runs. Two runs on one name at the same time are refused safely: the second delivers free and remembers nothing. If the memory cannot be read or saved, the run delivers free and says so.

Why does this Actor run with limited permissions?

Least privilege: it only needs its own storages and the named memory store it creates itself, so it asks for nothing else in your account.

More data tools by Automation Craft

• Meta Ad Library Scraper - All Placements, Filters
• Facebook Ads Library Scraper - Page Ads, No Login
• Instagram Ads Library Scraper - Creatives, Video
• TikTok Profile Scraper: Exact Stats, No Login
• LinkedIn Company Scraper - Details, No Login
• Google Trends Scraper - Compare and Trending Now

Changelog

• 0.1 (2026-10-07): first release.