Telegram Channel Scraper for LLM & RAG (No Login)
Pricing
from $0.60 / 1,000 posts
Telegram Channel Scraper for LLM & RAG (No Login)
Scrape posts from public Telegram channels without login or API keys: clean text, detected language, views, reactions, links and ready-made RAG chunks. No personal data stored. Pay only for posts delivered.
Pricing
from $0.60 / 1,000 posts
Rating
0.0
(0)
Developer
Henry Crawl
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
5 hours ago
Last modified
Categories
Features
Share
Scrape posts from public Telegram channels — no Telegram account, no API keys, no phone number — and get them back cleaned, language-tagged and ready for LLMs, embeddings and RAG pipelines.
It reads the public web preview Telegram serves to everyone at t.me/s/<channel>, pages back through the channel history newest-first, and returns one tidy row per post (or per RAG chunk).
Why this scraper
| This Actor | Typical Telegram scrapers | |
|---|---|---|
| Login / API keys | None | Often need a Telegram account, api_id/api_hash or session string |
| Text | Clean plain text, line breaks kept, emoji kept, links extracted | Raw HTML or markup |
| Language | Detected per post (language) — filter with languages | – |
| RAG | Optional chunks output with token estimates and stable ids | – |
| Filters before billing | Keywords, exclude words, language, minimum length, date, forwarded posts — filtered posts are free | Pay first, filter later |
| Personal data | No admin signatures, no hidden forward senders, no commenters | Often included |
| Errors | A free row explaining why a channel can't be read | Silent empty dataset |
Use cases
- News & market monitoring — follow crypto, finance, geopolitics or industry channels daily with Only posts since.
- RAG knowledge bases — turn a channel's history into embedding-ready chunks with
postId,dateandurlfor citations. - AI agents — give an LLM the latest posts of a channel as context ("what did @channel announce this week?").
- Research & OSINT on public channels — multilingual: Russian, Ukrainian, Persian, Arabic, Vietnamese, Chinese and more.
- Brand & competitor tracking — keyword filters across many channels in one run.
Input
{"channels": ["durov", "https://t.me/bbcrussian", "@telegram"],"maxPostsPerChannel": 200,"postsSince": "2026-11-01","includeKeywords": [],"languages": [],"outputFormat": "posts"}
- channels —
name,@nameort.melinks. Only public channels with web preview. - maxPostsPerChannel — newest first; the Actor pages back until this number or
postsSince. - includeKeywords / excludeKeywords / languages / minChars / includeForwarded — applied before billing.
- outputFormat —
posts(one row per post) orchunks(one row per RAG chunk;chunkTokens,chunkOverlapTokens). - includeChannelInfo — adds one free row per channel: title, description, subscribers, verified.
Output
One row per post (recordType: "post"):
{"recordType": "post","status": "OK","id": "durov-559","channel": "durov","channelTitle": "Pavel Durov","postId": 559,"url": "https://t.me/durov/559","date": "2026-11-13T09:29:23Z","edited": false,"text": "Founders — don't have a board. Be the board 😏","language": "en","views": 558000,"reactions": 48950,"forwarded": false,"forwardedFromChannel": null,"replyToPostId": null,"photos": 0,"videos": 0,"hasDocument": false,"pollQuestion": null,"links": [],"linkPreview": null,"hashtags": [],"charCount": 251,"wordCount": 47,"tokenEstimate": 63,"contentHash": "3f1c0d9a7b2e4c11","scrapedAt": "2026-11-13T14:30:00Z"}
With outputFormat: "chunks" each row is recordType: "chunk" with id (<channel>-<postId>-<chunkIndex>), text, tokenEstimate, chunkIndex, chunkCount and the post's metadata.
Error rows (recordType: "error", free): NO_PUBLIC_PREVIEW (channel doesn't exist, preview disabled, or it's a group / bot / person), NO_POSTS, INVALID_CHANNEL, RATE_LIMITED, HTTP_ERROR, NETWORK_ERROR.
A run summary is saved to the key-value store as SUMMARY.
Pricing
Pay per event: $0.60 per 1,000 posts delivered (plus Apify's tiny per-run start fee).
Filtered-out posts, channel info rows and error rows are free. In chunks mode you still pay once per post, not per chunk.
Limits & notes
- Works only for public channels whose owner left the web preview on (most news, media and project channels). Private channels, groups, invite links and personal accounts are out of scope — by design, no login is ever used.
- Telegram does not say whether an unavailable name doesn't exist or just has no preview; both return
NO_PUBLIC_PREVIEW. - Views and reactions are Telegram's rounded counters (e.g.
4.69M→ 4690000). - Media files are not downloaded; you get counts and the post URL.
- Requests are paced politely (one page at a time per channel).
- Admin signatures, the names of hidden forward senders and comments are deliberately not collected.
FAQ
Do I need a Telegram account? No.
How far back can it go? As far as the channel's public history goes; set maxPostsPerChannel high (up to 20,000) or use postsSince.
Can I monitor channels daily? Yes — schedule the Actor with postsSince set to yesterday and dedupe on id.