Telegram Channel Scraper — public posts and paginated history

Collect the posts exposed by public Telegram channel previews at t.me/s. Export visible post text, UTC publication dates, links, media references and available views and reactions into a Dataset for CSV, JSON and API workflows. No Telegram account, phone number, bot token or external API subscription is required.

Use it for channel research, editorial monitoring, link discovery and analysis of your own public channels. Older pages are followed using Telegram's real before cursor. This is a public preview collector: source gaps, deleted posts and restricted previews prevent a promise of complete history.

Quick start: five posts

{"channels":["telegram"],"maxPostsPerChannel":5,"maxResults":5}

Run the Actor, open Post Dataset, and export CSV or JSON. Open SUMMARY, COVERAGE, ERRORS and CHANNELS in the run's key-value store to understand what was scanned and why collection stopped. Missing data is represented by null or an empty collection, not an invented zero.

Price: $0.05 per 1,000 useful delivered posts on every plan, plus Apify resources. There is no startup event, channel profile event or filtered-post event. Resource use still costs money when no post matches. Set both the run spending limit and reasonable page/result/time limits.

What this Telegram scraper collects

Capability	Result and boundary
Multiple public channels	Handles, @handles and public channel URLs; equivalent inputs are deduplicated
Public post history	Latest visible posts, then older preview pages; newest post ID first within each channel
Text and links	Visible text with line breaks; HTTP/HTTPS links in the post body
Publication date	ISO UTC date from the post permalink's time element, separate from video duration
Views and reactions	Raw source labels plus approximate numbers; custom emoji identifiers retained when exposed
Media	Public photo/video/audio references and thumbnails when exposed; files are not downloaded
Filters	Inclusive UTC date range and case-insensitive keyword OR filter on visible text
Continuation	Per-channel nextBefore in COVERAGE; copy into a separate single-channel run
Diagnostics	Coverage, profile context and source errors outside the paid post Dataset

Useful inputs

Compare channels

{"channels":["telegram","https://t.me/s/durov"],"maxPostsPerChannel":50,"maxResults":100,"maxConcurrency":2}

Each row contains a channel identity. Concurrent channels can interleave; sort by channel/postId or publishedAt after export. Post IDs determine traversal order; dates are output and filter values, not a guarantee that every source date is monotonically ordered.

Find posts containing any phrase

{"channels":["telegram"],"keywords":["update","privacy"],"maxPostsPerChannel":20,"maxPagesPerChannel":25}

Filters apply to the pages actually scanned. A rare keyword can produce fewer rows or no rows. This does not prove the phrase is absent from all channel history.

Read a date range

{"channels":["durov"],"dateFrom":"2026-01-01","dateTo":"2026-12-31","maxPostsPerChannel":100,"maxPagesPerChannel":30}

dateTo without a time includes that whole UTC day. Old date ranges may require a known cursor or more pages. This version scans backward; it does not perform a binary date search.

Continue older history

{"channels":["telegram"],"beforePostId":400,"maxPostsPerChannel":100,"maxResults":100}

The numeric cursor is illustrative; use your channel's COVERAGE.nextBefore. It is exclusive: the previous last scanned/delivered post is not requested again. Continue channels separately because their IDs are independent. Retain (channel,postId) in your own database and deduplicate across separate runs. The Actor does not maintain a persistent cross-run monitoring state or guarantee exactly-once delivery across crashes.

Increase collection with explicit caps

{"channels":["telegram","durov"],"maxPostsPerChannel":1000,"maxResults":2000,"maxPagesPerChannel":100,"maxConcurrency":2,"maxRunSeconds":1800}

The accepted upper bounds are 1,000 channel inputs, 10,000 delivered posts per channel, 100,000 posts per run, 500 pages per channel and five concurrent channels. These are safety bounds, not measured throughput or source availability promises. A short channel can end before the requested count. Start small, inspect coverage and costs, then raise caps.

All settings

Input	Default	Meaning
channels	telegram at runtime	1–1000 public usernames or channel URLs; UI prefilled with telegram
keywords	[]	Up to30 nonempty phrases, max200 characters each; any phrase matches visible text
dateFrom	unset	ISO date/time, inclusive; timezone omitted means UTC
dateTo	unset	ISO date/time, inclusive; a date alone includes the whole day
beforePostId	unset	Exclusive older-history cursor; only one valid channel
maxPostsPerChannel	100	Delivered post cap per channel, 1–10000
maxResults	1000	Shared delivered post cap, 1–100000
maxPagesPerChannel	100	Source page cap per channel, 1–500
maxConcurrency	3	Channels in parallel, 1–5; pages within a channel remain sequential
maxRunSeconds	900	Collection budget, 10–3300 seconds, checked between requests

Set the Apify run timeout above maxRunSeconds to leave time for in-flight requests, retries and final storage. Defaults use 512MB and a 3600-second Actor timeout. Requests use bounded retries for network failures, 429 and transient server errors. Repeated pages and non-advancing cursors stop collection with diagnostics instead of looping.

Output fields

Field	Meaning
channel, channelTitle	Normalized public username and current preview title
postId, postUrl	Numeric post ID and public permalink; identity is channel plus ID
publishedAt	Source publication timestamp in UTC; this is not the scrape time
text	Visible post text, nullable for media-only posts
links	Deduplicated HTTP/HTTPS links inside the visible post text
media	Available public photo/video/audio references, videoThumbnail, or documentPreview metadata; no file download. documentPreview has title/description/postUrl and url=null because the source exposes no file URL
mediaUnavailable	Telegram labels some media as unsupported/restricted; may be true even if a thumbnail is exposed
viewsLabel, viewsApprox	Source label, e.g. 1.29M, and approximate numeric conversion
reactions	Exposed emoji/customEmojiId/isPaid/countLabel/countApprox items; empty if none exposed
metricsAreApproximate	Always true: rounded source metrics are not exact counters
source, scrapedAt	Public-preview provenance and collection time

Reaction labels, custom emoji and paid/star reactions differ between channels. A custom emoji ID does not imply a known Unicode emoji. A profile's rounded subscriber count is stored in CHANNELS. No channel member list, personal contact information, commenter profiles or phone numbers are collected.

SUMMARY records delivered rows, fetched pages, filtered/invalid rows, duplicates, errors and elapsed time. COVERAGE records each channel's page count, stop status and continuation cursor. SOURCE_END means no older link was exposed in this traversal; it does not prove complete historic coverage. PAGE_LIMIT, CHANNEL_RESULT_LIMIT, GLOBAL_STOP and BUDGET_LIMIT indicate bounded collection. Source errors remain in ERRORS. A mixed run may succeed with useful posts and errors; inspect diagnostics. A source-only failure without posts fails the run, while a valid filter producing no matches succeeds.

Billing and costs

Only useful unique post rows delivered through the SDK's coupled Dataset/event operation use the post event. Duplicate, filtered, invalid-input and diagnostic records are not post events. Channel context is not a separate paid event. An uncertain Dataset/billing failure fails the run; it is not silently retried.

The event component is delivered posts × $0.00005: 100 posts cost $0.005 in events; 1,000 cost $0.05. Add the actual Apify platform usage shown on the run. This is not an all-inclusive price or a fixed total quote. HTML size, scanned pages, retries and filters change resource cost. No proxy is configured or required by this version; source blocking can still occur. Configure maxTotalChargeUsd when launching through the API or the Console's spending limit.

API, Python and automation

Call the Actor using apivault_labs~telegram-public-channels-scraper and your own Apify token. A private candidate is accessible only to its owner until publication. Store tokens as credentials, never in shared input examples.

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("apivault_labs~telegram-public-channels-scraper").call(
    run_input={"channels": ["telegram"], "maxPostsPerChannel": 5},
    max_total_charge_usd=0.10,
)
for post in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(post["channel"], post["postId"], post["publishedAt"])

Use iterate_items() for paginated Dataset reads. One API request with a default response limit is not a full Dataset download. Official Python client and Apify API reference describe token authentication, runs, Dataset export and key-value records. n8n or Make can call the same API through standard HTTP request nodes and stored credentials. No Telegram-specific verified n8n node, custom Python package or MCP execution is claimed. This pass provides input examples rather than new public Tasks.

Limits and troubleshooting

Only publicly exposed channel previews are supported. Private groups, invite links, personal profiles, member lists, login-only content, comments and full reply trees are excluded. Forward counts and complete media downloads are not promised. A post may have text plus an unavailable-media warning; the warning is retained. Signed media links can expire or stop working independently of post URLs.

Preview-disabled, removed or renamed channels can redirect or return no preview. These become source errors; they are not fake empty successes or paid placeholder posts. Public groups are rejected when the source identifies members rather than channel subscribers. Rate limits and source markup changes can interrupt history collection. The Actor does not bypass access controls.

Why fewer posts than requested? The source ended, a filter excluded posts, a page/time/output/budget cap was hit, or a source error occurred. Read COVERAGE and ERRORS.

Are views, reactions and subscriber counts exact? No. Source labels such as K/M are rounded. Preserve the labels when reporting metrics.

Is this real-time monitoring? Each run is a snapshot. You can schedule it yourself and deduplicate IDs in your database; persistent monitoring/baselines are not built in.

Can I collect the complete archive? Only the older pages exposed by the public preview. Deleted/restricted posts and source gaps remain unavailable.

Python client and n8n workflow

Use the thin Python client or install it from PyPI:

$pip install apivault-telegram-public-channels-scraper==0.1.0

The client starts this hosted Actor through the Apify API and reads its dataset. An Apify account and your own API token are required. Actor result charges and infrastructure usage follow the current pricing. The client package has no subscription of its own. Set both Actor collection limits and run charge caps; the dataset download limit does not limit paid collection.

n8n node source and manual workflow are available on GitHub. The workflow is inactive and has no embedded credentials. Its Run operation starts a paid Actor run; Read Existing Run reads a previous run without starting another. npm version0.1.0 currently awaits npm staged-release approval; do not install the placeholder0.0.0-stage as an operational node. npm publication, when approved, does not establish n8n verification or availability in n8n Cloud.