Lemmy Posts, Comments, Threads & Communities (no login)

Public Lemmy content as flat rows, read from the HTTP API of lemmy.world — the largest instance of the network — which answers JSON without an account, a token or a cookie. Six questions in one actor: the posts of the communities you name, the site-wide feed, a keyword across posts and comments, the whole discussion under a post numbered by reply depth, what an account published in public, and which communities exist for a topic with their subscriber and activity counts. Lemmy is federated, so that one server also holds content from thousands of other instances: every row names the server its author and its community live on, and every timestamp is UTC ISO-8601.

Need it tailored to your workflow? Missing a field, a filter, another source or a different output format? Open an issue and describe what you need — we read every request and adapt the actor.

Use cases

• Keep a topic feed after a forum closes its own. Point communityPosts at technology, asklemmy or any community, set sinceHours to your schedule interval and onlyNew: true — each run then holds exactly what appeared since the last one.
• Watch a product name where opinions actually are. searchPosts finds the posts, searchComments the replies; on Lemmy most of what people think about a release sits in comments, and both modes reach communities hosted on other instances.
• Summarise a discussion. postComments turns one post link into its thread in reading order, with replyDepth, parentCommentId and childCount, which is the shape an LLM needs to answer "what did people reply, and to whom".
• Archive public announcements. userActivity exports what a moderator, admin or project account posted and commented, with the lifetime post and comment totals the profile shows.
• Size the network before you monitor it. communities ranks communities for a topic by subscribers and users active per day, week and month, so a listening pipeline starts with the communities that are alive.
• One feed for a dashboard. communityPosts with no community named is the site-wide feed; listingType: local keeps it to the communities lemmy.world hosts itself, all opens it to the federated view.
• Study federation. authorInstances and communityInstances filter on the server columns — watch one instance's voices inside a worldwide community, or drop an instance that floods a feed.

Input

Every field is optional and each mode uses the fields of its own section. Values shown in the editor are prefills — change or clear them freely; only behaviour settings have real defaults.

Field	Type	Default	Allowed values / notes
mode	string	communityPosts	communityPosts, searchPosts, searchComments, postComments, userActivity, communities — see Modes
communities	string[]	prefill technology, asklemmy	Communities to read. name, name@server or a community link — see Writing a community or a handle. Empty in communityPosts = the site-wide feed
listingType	string	all	all = every federated server this one knows, local = communities hosted on lemmy.world. Applies to the site-wide feed, both search modes and the community directory
sort	string	New	16 orders for posts, comments and communities — see Sort orders
searchQuery	string	prefill kubernetes	The keyword for searchPosts, searchComments and the community directory. One phrase per run, matched as a case-insensitive substring
postUrls	string[]	prefill one post link	postComments: a lemmy.world link, a bare post id, or a post link on any other instance — see Post links
includeOriginalPost	boolean	true	postComments: write one post row per thread before its comments
maxCommentDepth	number	8	1–50. Levels of replies to read in postComments; 1 = direct replies only
commentSort	string	Top	Top, New, Old, Hot, Controversial — the order of the branches in postComments
users	string[]	prefill ruud@lemmy.world	userActivity: accounts as name, name@server or a profile link. The row limit is split over them
userContent	string	posts	userActivity: posts, comments or both (split evenly, newest first)
languages	string[]	all	Keep rows whose declared language is one of 22 codes, und included — see Languages
authorInstances	string[]	all	Keep rows whose author lives on one of these servers, e.g. lemm.ee, programming.dev
communityInstances	string[]	all	Keep rows whose community is hosted on one of these servers
minScore	number	–	−100 000…100 000. Keep posts and comments with at least this score (upvotes minus downvotes)
minComments	number	–	0–100 000. Keep posts with at least this many comments. Comment rows carry childCount instead
excludeNsfw	boolean	true	Drop posts, comments and communities flagged NSFW. Switch off for a complete sample
excludeBots	boolean	false	Drop rows of accounts that declare themselves bots (authorIsBot)
sinceHours	number	–	1–8760. Rolling window in UTC hours. With sort: New paging also stops as soon as the source walks past it
onlyNew	boolean	false	Remember the post and comment ids written — as they are written — and write only unseen ones on the next run of the same actor or task
maxItems	number	50	1–1000. Hard cap on rows and therefore on cost; split over the communities, users or threads of the run. The run also stops at your maximum cost per run
fields	string[]	all	Keep only these columns, in this order; rowType, sourceMode, found and url are always kept. Letter case is corrected, unknown names are reported in the status message with the nearest column, and a list with no valid column fails the run

Reference

Modes

mode	Endpoints used	Rows	Dataset view
communityPosts	post/list	one post per post of the named communities, or of the whole instance	Posts
searchPosts	search (Posts)	one post per match in a title, body or posted link	Posts
searchComments	search (Comments)	one comment per match in comment text	Comments
postComments	resolve_object → post → comment/list	one post row (optional) plus one comment per reply, in reading order	Comments
userActivity	user	one post or comment per public item of the named accounts	Posts / Comments
communities	community, search (Communities), community/list	one community per community, with counts	Communities

Each mode fills one view; the others stay empty. communities answers the communities you name (with moderatorCount), otherwise the matches for searchQuery, otherwise the ranked directory for sort and listingType.

Sort orders

New, Old, Hot, Active, Scaled, MostComments, NewComments, Controversial, TopHour, TopSixHour, TopTwelveHour, TopDay, TopWeek, TopMonth, TopYear, TopAll.

A Top* order is a window as well as an order: TopWeek can never return a post older than seven days, so do not combine it with a longer sinceHours and expect more rows. New is the one order that walks strictly down the clock, which makes it the cheapest choice for a monitoring run — it is the only one where a window can end the paging early. Active and NewComments rank by recent comment activity, Scaled lifts smaller communities, Controversial surfaces what was up- and downvoted alike. Comment branches use the shorter list: Top, New, Old, Hot, Controversial.

Writing a community or a handle

A bare technology means the community of that name on lemmy.world. name@server (programming@programming.dev, climate@slrpnk.net) means the federated copy this server keeps of a community hosted elsewhere. Links work too: https://lemmy.world/c/technology, https://programming.dev/c/programming, and the !name@server form people paste from chat. Accounts follow the same rules with /u/: ruud, ruud@lemmy.world, https://lemm.ee/u/sunaurus. A name this server has never seen yields one row with found: false and the reason in notFoundReason, so one typo in a long list never disappears silently. The authorHandle column is always written as name@server, so rows from different instances can be compared.

Post links

postComments accepts https://lemmy.world/post/920294, the bare id 920294, and a post link on any other instance such as https://programming.dev/post/57357662. Every instance numbers posts itself, so a foreign link is first resolved to the copy lemmy.world holds (one extra request), and the row then carries the local postId and url plus the canonical apId of the origin server. Any post row of any mode already carries postId and url, which is how a feed or search run chains into a thread run.

Languages

und (not set), en, de, fr, es, pt, it, nl, sv, da, nb, fi, pl, cs, ru, uk, tr, ja, zh, ko, ar, ca. Only these codes are accepted: Apify refuses any other value - a name like English, a three-letter code like eng - before the run starts, so a run never keeps a language you did not pick. Read the honest part first: Lemmy stores a language per post and comment, but most clients never set it, so a large share of real rows carry und. Pick en and und together if you want English-speaking communities without losing most of their content.

Row types

rowType is post, comment or community; sourceMode repeats the mode that produced the row, and found is false only on a row that explains an input the source would not serve.

Examples

Posts of one community, last two days, newest first:

{ "mode": "communityPosts", "communities": ["technology"], "sort": "New", "sinceHours": 48, "maxItems": 20 }

The week's discussions of a question community, only the ones that got answers:

{ "mode": "communityPosts", "communities": ["asklemmy"], "sort": "TopWeek", "minComments": 5, "maxItems": 15 }

One keyword across every community the server knows:

{ "mode": "searchPosts", "searchQuery": "docker", "sort": "TopMonth", "listingType": "all", "maxItems": 15 }

The same keyword in comments, newest first:

{ "mode": "searchComments", "searchQuery": "kubernetes", "sort": "New", "maxItems": 15 }

A whole thread, two levels deep, highest-scoring branches first, comments only:

{ "mode": "postComments", "postUrls": ["https://lemmy.world/post/920294"], "commentSort": "Top", "maxCommentDepth": 2, "includeOriginalPost": false, "maxItems": 20 }

Which communities exist for a topic, biggest first:

{ "mode": "communities", "searchQuery": "climate", "sort": "TopAll", "maxItems": 15 }

The site-wide feed of the instance itself, for a dashboard:

{ "mode": "communityPosts", "listingType": "local", "sort": "Active", "maxItems": 20 }

Output

One real row of the cloud run SYSbBnAaKNBhd6CbX (input {"mode": "communityPosts", "communities": ["technology"], "sort": "New", "sinceHours": 48, "excludeNsfw": true, "maxItems": 20}):

{
  "rowType": "post",
  "sourceMode": "communityPosts",
  "found": true,
  "notFoundReason": null,
  "postId": 52605462,
  "title": "Google rolls out new Gemini AI model but restricts access over safety concerns",
  "body": null,
  "postUrl": "https://www.theguardian.com/technology/2026/oct/01/google-releases-gemini-model-restrictions",
  "thumbnailUrl": "https://lemmy.world/pictrs/image/583dd762-f759-4431-901f-851b259e0901.jpeg",
  "embedTitle": "Google rolls out new Gemini AI model but restricts access over safety concerns",
  "embedDescription": "Tech company releases Gemini 4 Argon only to a vetted group of cybersecurity experts to avoid misuse by hackers",
  "community": "technology",
  "communityTitle": "Technology",
  "communityInstance": "lemmy.world",
  "authorHandle": "orwi@programming.dev",
  "authorDisplayName": null,
  "authorInstance": "programming.dev",
  "authorIsBot": false,
  "authorPostCount": null,
  "authorCommentCount": null,
  "score": -1,
  "upvotes": 2,
  "downvotes": 3,
  "commentCount": 2,
  "newestCommentAt": "2026-10-01T21:43:29.984Z",
  "createdAt": "2026-10-01T21:18:55.327Z",
  "updatedAt": null,
  "nsfw": false,
  "locked": false,
  "removed": false,
  "deleted": false,
  "featuredInCommunity": false,
  "featuredSiteWide": false,
  "language": "en",
  "matchedCommunity": "technology",
  "matchedQuery": null,
  "url": "https://lemmy.world/post/52605462",
  "apId": "https://programming.dev/post/57357662",
  "fetchedAt": "2026-10-01T22:28:58.531Z"
}

Download the dataset as JSON, CSV, Excel, HTML or XML, or read it over the API.

Post row (rowType: "post")

Field	Type	Meaning
postId	number	Id of the post on lemmy.world; the input of a postComments run
title	string	Post title as written
body	string	Post text, Markdown as the source stores it; null for a pure link post
postUrl	string	The link the post points at, null for a text post
thumbnailUrl	string	Preview image the server generated
embedTitle, embedDescription	string	Title and summary the server read from the linked page
community, communityTitle	string	Community name and its display name
communityInstance	string	Server hosting the community, e.g. lemmy.world, slrpnk.net
authorHandle	string	Author as name@server, always in that form
authorDisplayName	string	Chosen display name, often null
authorInstance	string	Server the author's account lives on
authorIsBot	boolean	The account declares itself a bot
authorPostCount, authorCommentCount	number	Lifetime totals of the account — filled in userActivity, null elsewhere
score, upvotes, downvotes	number	Public vote totals at fetchedAt; the score can be negative
commentCount	number	Comments the server counts under the post
newestCommentAt	string	UTC time of the newest comment
createdAt, updatedAt	string	Posted and last edited, UTC ISO-8601
nsfw	boolean	Flagged by the post or by its community
locked	boolean	Closed for new comments
removed, deleted	boolean	Moderator removal / author deletion flags the source publishes
featuredInCommunity, featuredSiteWide	boolean	Pinned in the community or on the front page
language	string	Declared language code, und when the client set none
matchedCommunity	string	The community of the input that produced the row
matchedQuery	string	The keyword that produced the row
url	string	Permalink on lemmy.world
apId	string	Canonical link on the instance the post was written on
found, notFoundReason	boolean, string	false plus a reason when the source would not serve the input
fetchedAt	string	UTC time this row was read

Comment row (rowType: "comment")

Field	Type	Meaning
commentId	number	Id of the comment on lemmy.world
text	string	Comment text, Markdown as written
replyDepth	number	1 for a direct reply to the post, 2 for a reply to that, and so on
parentCommentId	number	The comment this one answers, null at depth 1
childCount	number	Replies hanging below it — how much a depth or row limit cut off
postId, postTitle	number, string	The post the comment sits under
community, communityTitle, communityInstance	string	Where the discussion happens
authorHandle, authorDisplayName, authorInstance, authorIsBot	string, boolean	The comment's author
score, upvotes, downvotes	number	Public vote totals at fetchedAt
createdAt, updatedAt	string	Written and last edited, UTC
removed, deleted, nsfw	boolean	Moderation flags; nsfw is inherited from the post or community
language	string	Declared language code, often und
matchedQuery, matchedCommunity	string	What produced the row
url, apId	string	Permalink here and canonical link on the origin instance
fetchedAt	string	UTC time this row was read

Community row (rowType: "community")

Field	Type	Meaning
community, communityTitle	string	Name and display name
description	string	Sidebar text, Markdown
communityInstance	string	Server hosting it
subscribers, subscribersLocal	number	Subscribers everywhere, and from lemmy.world alone
postsCount, commentsCount	number	Totals the server counts
usersActiveDay, usersActiveWeek, usersActiveMonth, usersActiveHalfYear	number	Accounts that posted or commented in the window
moderatorCount	number	Moderators — filled for communities you name, null for search and directory rows
nsfw, postingRestrictedToMods	boolean	Adult flag; only moderators may post
createdAt, updatedAt	string	Created and last changed, UTC
icon	string	Community icon URL
matchedQuery	string	The topic word that produced the row
url, apId	string	Link here and canonical link on the origin instance
fetchedAt	string	UTC time this row was read

Views and the run summary

Three dataset views pick the columns worth a table: Posts (author, community and its server, title, score, comments, date, link), Comments (author, post, text, score, depth, replies below) and Communities (server, subscribers, active users, posts, comments, moderators). Every run also writes a SUMMARY record into its default key-value store with the request count, the rows per community, user or thread, the skipped rows per filter, the inputs the source would not serve and every warning — useful when a scheduled run returns fewer rows than expected.

Use it from code / agents

curl -X POST "https://api.apify.com/v2/acts/yadroo~lemmy-public-posts/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"mode":"searchComments","searchQuery":"kubernetes","sort":"New","maxItems":20,"fields":["authorHandle","community","postTitle","text","score","createdAt","url"]}'

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('yadroo/lemmy-public-posts').call({ mode: 'communityPosts', communities: ['technology'], sinceHours: 24, maxItems: 50 });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const row of items) console.log(row.createdAt, row.authorHandle, row.score, row.title);

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
# 1) find the busiest post of a community, 2) read its discussion for a summary
feed = client.actor("yadroo/lemmy-public-posts").call(
    run_input={"mode": "communityPosts", "communities": ["asklemmy"], "sort": "TopWeek", "maxItems": 20}
)
rows = client.dataset(feed["defaultDatasetId"]).list_items().items
busiest = max(rows, key=lambda r: r.get("commentCount") or 0)
thread = client.actor("yadroo/lemmy-public-posts").call(
    run_input={"mode": "postComments", "postUrls": [busiest["url"]], "commentSort": "Top", "maxCommentDepth": 3, "maxItems": 100}
)

Chaining works in both directions: a communities run gives community and communityInstance values to paste straight into communities (name@server), and any post row gives a url for postComments. For a schedule, set sinceHours a little above the interval and onlyNew: true — an empty run then means nothing new was posted. Narrow rows with fields when an LLM reads them.

MCP: add https://mcp.apify.com to Claude, Cursor or any MCP client and call the yadroo/lemmy-public-posts tool with the same JSON input.

Pricing

Pay per event: $0.001 per run start + $0.0015 per dataset row. Platform usage (compute and traffic) is included in these prices — nothing is charged on top. The start event is charged on every run, including a run that writes no row (a monitoring run with onlyNew and nothing new, for example). A found: false row — the one row that explains an unknown community, account or post — is a dataset row and is billed like any other; a post or comment the source serves twice in one run is written and billed once.

Run	Rows	Cost
Community directory for a topic	15	$0.001 + 15 × $0.0015 = $0.0235
Community feed, last 24 h	50	$0.001 + 50 × $0.0015 = $0.0760
A busy thread, three levels deep	200	$0.001 + 200 × $0.0015 = $0.3010

Store discounts lower the row price automatically: Bronze −10 %, Silver −20 %, Gold and above −30 %; the start event is the same on every plan. Set Maximum cost per run and the run stops exactly there: it writes the rows that budget pays for and ends with "Stopped at your spending limit: N rows delivered". Keep maxItems small while you try inputs out — it is the one input that decides what a run costs.

Limits & FAQ

• 50 rows per request, so deep runs page. The source refuses limit above 50 with an HTTP 400, which the actor clamps for you; a 1000-row run is roughly 20–25 requests and a few minutes.
• Politeness and rate limits. lemmy.world publishes its own per-IP limits (100 search calls per 10 minutes, 999 other reads per minute). The actor sends requests one at a time with a pause of ~350 ms, spaces searches after the first ten to one every 6 seconds, retries HTTP 5xx with backoff and honours Retry-After. Apify runs share outgoing IPs, so the search budget can already be partly used when a run starts: when the server refuses a request for its limit (HTTP 400 rate_limit_error, or 429), the actor waits 30 s, then 60 s, and if the refusal persists it stops reading and says so in the status message — the rows written so far are kept, a run that wrote nothing fails, and the next run should start about 10 minutes later. It never switches IPs to get around the limit. Only /api/v3/* of that one host is ever called, never the HTML pages.
• Spending limit, timeout, restarts. Rows are written page by page as they are read. The run stops at your maximum cost per run ("Stopped at your spending limit: N rows delivered"), stops starting new requests shortly before the run timeout and ends successfully with what it saved ("Stopped before the run timeout: N rows saved"), and a platform migration or restart never writes or bills a row twice.
• onlyNew memory. Rows are remembered as they are written, in a named key-value store of your account (lemmy-seen, or lemmy-seen-<task id> for a task, up to 50 000 ids), so a run cut short by maxItems, the spending limit or the timeout leaves the rest for the next run. A found: false row is remembered too: an unknown name gives one row once, and afterwards appears only in the status message.
• A thread request is capped at 300 comments. Asked for a depth limit, the source answers up to 300 comments of a post and ignores paging there; the actor then continues with the pageable form of the same endpoint to fill maxItems, and the status message says how many of the thread's comments it could read. childCount shows where a branch was cut.
• No URL or domain search. The source accepts a URL search type but answered zero posts for domain queries, so "who linked my site" is not offered rather than offered broken.
• No phrase or boolean search. searchQuery is one case-insensitive substring: no quotes, no AND/OR, no wildcards. For several phrases, run several tasks and merge the datasets.
• sinceHours is applied here, not by the source. There is no date parameter in the API. With sort: New the actor stops paging at the first row past the window; with any other order the window only filters what the chosen ranking already returned.
• Most content declares no language. language is und for a large share of rows. The languages filter can only keep what declares a language — add und to the list or leave the filter empty.
• A federated view, not all of Lemmy. Rows are what lemmy.world holds: content of instances it federates with, and nothing from instances it has defederated. communityInstance and authorInstance tell you where each row really comes from, and a post can exist elsewhere on the network without existing here.
• Vote totals are a snapshot. score, upvotes and downvotes are the values at fetchedAt and keep moving; votes are only ever read as the totals the server publishes, never per person.
• Nothing behind a login, no contact harvesting. Only what a signed-out visitor sees: posts, comments, pseudonymous handles, public vote totals, community counts and the post/comment totals a profile shows. No follower or subscription lists, no e-mail addresses, no chat identifiers — the source publishes a chat handle on some profiles and this actor deliberately does not write it. Images are linked, never downloaded.
• Empty result vs. error. A quiet community, a keyword nobody used or a filter that dropped everything is a successful run with the reason in the status message and in SUMMARY. An unknown community, account or post gives one row with found: false and a reason. A run that wrote nothing and hit a misbehaving source — or was stopped by the server's rate limit — fails with that reason, so a schedule can tell the two apart.
• Deleted and removed content. The API already leaves it out; the removed and deleted columns stay in the row shape for the cases where it still answers with a flag.
• If the server closes these endpoints. A Lemmy instance can put its API behind authentication. Should that happen here, this actor is marked paused on this page instead of working around a block.

---

Made by Yadroo. Sibling actors: mastodon-public-posts, bluesky-public-posts, hackernews-search, stackexchange-search, rss-to-json.