Telegram Keyword Search Scraper | Brand Mentions
Pricing
from $1.00 / 1,000 telegram posts
Telegram Keyword Search Scraper | Brand Mentions
Search public Telegram posts by keyword or brand with AND/OR/NOT operators and view, subscriber, media and language filters. Export CSV/JSON, no Telegram login. $1/1K posts + $0.005/verified discovery page + start fee. Telegram关键词搜索与品牌监测,无需登录。
Pricing
from $1.00 / 1,000 telegram posts
Rating
5.0
(1)
Developer
Monkey D. Scraping
Maintained by CommunityActor stats
3
Bookmarked
56
Total users
18
Monthly active users
10 hours ago
Last modified
Categories
Features
Share
Find public Telegram posts that mention a brand, product or topic. Start with a keyword, choose Post text only and review each match against its original Telegram post. Export post text, dates, channels and source links to CSV, Excel or JSON. No Telegram login or API key required.

Try a five-post keyword search → Replace the example keyword, keep the small result limit, and review the maximum run cost before starting.
中文用户 / 中文用戶: 简体中文说明 · 繁體中文說明(Telegram关键词搜索、Telegram品牌提及与舆情监控、导出CSV/JSON)
- Check the mention itself: match post text and skip recognized signature-only matches, rather than relying on the channel title.
- Compare sources: limit posts per channel or turn on source sampling to reduce concentration in the first productive channel.
- Control the spend: set a maximum run cost independently of the post limit.
- Repeat a keyword search: open the repeat input from the result guide to keep your filters and skip posts already delivered. The follow-up searches your matched channels without another discovery page fee.
Price: $0.001 per saved post + $0.005 per verified discovery page + $0.00005 per start at up to 1 GB. Platform usage is included. One verified page and five saved posts cost $0.01005. A search with no results costs only the start event: confirmed empty discovery pages are not charged.
Example follow-up: a Bitcoin search found one post in a public channel. The follow-up kept the same keyword, skipped that post and returned another from the same channel. At the listed event prices, one follow-up post costs $0.00105 at up to 1 GB, including the start event. See how to repeat the search.
The linked Quick starter checks one discovery page for one keyword and has a saved $0.02 run cap. The main Actor prefill uses Standard coverage and can check more pages. Each result is a literal keyword mention with its original post link, and Results and next steps summarizes every run.
Already have channel URLs? Use Telegram Channel Scraper for public channel history and CSV export without keyword discovery. For repeated collection with checkpoints, see the reuse section below before choosing a monitoring workflow.
Choose speed, more matches or more sources
The starter checks post text, deduplicates saved posts within the run, and keeps original source links. These controls serve different research goals.
| Goal | Settings to adjust | Effect |
|---|---|---|
| Small first sample | Keep one keyword, five posts and Quick search coverage | Checks one discovery page per keyword |
| More matches from discovered channels | Keep Search inside discovered channels enabled; raise the total/per-keyword limits only as needed | Bounded extra checks can recover additional matching posts, including from channels already producing matches |
| Less concentration in one channel | Set Maximum posts per channel to 2 | Limits dominance by a single channel; can require more discovery |
| Earlier stop on sparse searches | Under advanced search limits, set Stop keyword after pages without new matches to 1 | Stops after the first verified page with no new saved matches; the default is 2 |
Start with one keyword and set Maximum cost per run separately. A post limit does not cap discovery charges. Keep the default stop of 2 when recovering later matches matters more than reducing work. Country and language are optional discovery preferences; keywords are searched as written.
Keyword operators
One keyword line is one literal phrase. To combine terms inside one line, use AND, OR, NOT (or a leading -) and "quoted phrases":
lavazza AND capsule OR cialde NOT recensione"a modo mio" -sconto
A post matches when every AND group has at least one literal term in the text and no excluded term appears. The matched keyword is reported as the whole line, so each line stays one countable topic. Discovery uses the required terms; exclusions apply to the saved posts.
Quality filters
Under Quality filters you can keep only posts with at least a given public view count (minViews), from channels with at least a given public subscriber count (minSubscribers), skip forwarded posts (excludeForwards), require or exclude attachments (mediaFilter) and keep only detected post languages (postLanguages, ISO 639-1). Language is a local statistical estimate on posts of at least 20 characters, reported as detectedLanguage; shorter posts have null and are skipped only when the language filter is set. These filters apply before saving and never add discovery charges; rejected counts appear in SUMMARY.filterDiagnostics and in the result guide.
Check whether broader recovery helped
Compare unique post links, distinct channelUsername values, the actual text and total run cost. SUMMARY.channelSearch.postsSaved counts matches saved by internal channel search. Deduplication applies within each run; the repeat input below skips posts already delivered.
Observed volume-versus-sources check
In two publisher-run checks on 5 November 2026, the same five queries (Lavazza, NotebookLM, energia rinnovabile, Apple and a nonsense negative control) used Standard coverage, US/en discovery preferences, internal channel search, an early-stop setting of 1, ten posts per keyword and fifty total. Only the per-channel cap changed; both runs had a $0.09 spending cap.
| Configuration | Unique posts | Distinct channels | Console duration | Event total |
|---|---|---|---|---|
| No per-channel cap | 37 | 7 | 62 seconds | $0.06705 |
| At most 2 posts per channel | 35 | 20 | 124 seconds | $0.08005 |
On 8 November 2026 the same five queries, with no per-channel cap, ran on the parallel release candidate: 35 unique posts from 13 channels in 23 seconds, first post saved after 2.5 seconds, $0.07005 for seven verified discovery pages and 35 posts. Four of the five queries finished in under 4 seconds each. Discovery now checks post-level public views before the web index, reads channels in parallel and keeps every saved post a literal match.
All saved posts contained a query in their body text, and the negative control returned zero in both. These are two observed samples; public discovery changes over time. Limiting concentration took more work and returned posts from more channels. Prices shown include the observed verified discovery pages, posts and one start at 128 MB.
Choose your goal: ready-to-run tasks
Start with the task closest to your goal. Each task works with public Telegram channels, require no Telegram login, and export one row per saved post. Save your own copy, review Run options, then start it.
| Goal | Ready-to-run task | Saved example | Expected result | Example event total |
|---|---|---|---|---|
| Find brand or product mentions | Find brand mentions → CSV | NotebookLM; Standard discovery; last 90 days; at most two posts per channel | Up to five posts with literal mentions in post text, dates and source links | $0.02005 for three verified discovery pages and five posts; saved run cap $0.03 |
| Research a topic | Research a topic → keyword starter | bitcoin; Quick discovery; at most two posts per channel | Up to five text matches across available sources, with source links | $0.01005 for one verified discovery page and five posts; saved run cap $0.02 |
| Export recent channel posts | Export recent channel posts → CSV | Three public channels found by the topic starter; last seven days; at most one post per channel | Up to three recent posts with no keyword discovery | $0.00305 for three posts; saved run cap $0.005 |
| Compare competitor mentions | Telegram Competitor Mentions: Perplexity vs ChatGPT | Perplexity and ChatGPT; Standard discovery; English; last 30 days; up to five posts per keyword and two per channel | Up to 10 posts that mention either brand, with text, date, channel and post link | Saved run cap $0.06 |
Browse all ready-to-run examples → — English, 中文 and Русский.
Примеры на русском: Поиск в Telegram по ключевому слову: нейросеть
These are standard event-price calculations at up to 1 GB memory, including platform usage: $0.001 per saved post + $0.005 per verified discovery page + $0.00005 per start. A run cap is a spending limit, not a fixed fee. Confirmed empty discovery pages are not charged; fewer posts reduce the post charge. Check the live Pricing tab before running.
1. Find brand mentions
Replace NotebookLM with your brand or product name. Each row is a literal mention with its source link, so you can open the original post.
{"collectionMode": "keyword","keywords": ["NotebookLM"],"channels": [],"maxResults": 5,"maxResultsPerChannel": 2,"afterDate": "90 days","discoveryCoverage": "standard","keywordMatchScope": "postText","includeContextPosts": false,"downloadMedia": false,"searchWithinChannels": false}
2. Research a topic
Replace bitcoin with a focused topic. Use words in the language of the posts you want to find; translations are separate queries and can add discovery charges. Start with one keyword.
{"collectionMode": "keyword","keywords": ["bitcoin"],"channels": [],"maxResults": 5,"maxResultsPerKeyword": 5,"discoveryCoverage": "quick","maxResultsPerChannel": 2,"spreadAcrossChannels": true,"keywordMatchScope": "postText","excludeSignatureOnlyMatches": true,"includeContextPosts": false,"downloadMedia": false}
3. Export recent channel posts
Start with the public channels found in the topic starter, or replace them with your own sources. This collects recent posts without keyword discovery. For a new keyword search, go back to the first two tasks. For incremental collection with a checkpoint, use Telegram New Posts Monitor.
{"collectionMode": "channel","keywords": [],"channels": ["bitcoin", "bitcoin_industry", "bitcoinnews"],"afterDate": "7 days","maxResults": 5,"maxResultsPerChannel": 1,"downloadMedia": false}
Observed starter and channel follow-up
On November 6, 2026, the topic starter saved five bitcoin matches from three channels in 10.4 seconds for $0.01005. The generated channel input then saved three recent posts for $0.00305, with zero discovery pages. Two had appeared in the first run: the follow-up collects recent channel activity.
Example source links from the follow-up: bitcoin/20716, bitcoin_industry/2439, and bitcoinnews/16393.
From input to a useful export
- Open a task above or paste its JSON into a new Actor run. Set the corresponding maximum cost in Run options; these JSON examples do not carry a spending limit.
- After the run, open Results and next steps for a readable summary of the run.
- Review the dataset fields below and open the original post links. Export CSV for a spreadsheet or JSON for an API workflow.
- For n8n, Make or an MCP client, pass the same input keys, wait for the run to finish, then read the default dataset and
SUMMARY. Choose keyword or channel mode explicitly. Use Telegram New Posts Monitor when you need repeated collection with a checkpoint.
| Output field | What to use it for |
|---|---|
text | Review the actual mention or post; can be null for textless posts |
messageUrl | Open and verify the original public post |
publishedAt | Sort by publication time; null when not exposed |
channelUsername | Identify the source and reuse it in a channel-only run |
matchedKeywords / matchType | Inspect how the post matched; channel-only collection is not a keyword match |
Illustrative output excerpt:
{"channelUsername": "example_channel","messageId": "123","messageUrl": "https://t.me/example_channel/123","publishedAt": "2026-11-01T12:00:00.000Z","text": "Example post mentioning NotebookLM."}
Observed product-mentions example
Try the product-mentions example task. Open your own copy with NotebookLM, a 90-day window, post-text-only matching and up to five posts. The saved run cap is $0.03; check Run options before starting. Review the original posts, then export CSV.
Start with the examples below, then export your dataset as JSON or CSV from the run page.
By Monkey D. Scraping. Start with a product name when you do not know which public channels to check. Select Post text only so saved keyword matches contain the term in their text. Each row links to the original post, so you can read the mention in context.
A publisher-run NotebookLM example on October 30, 2026 returned five posts from three channels in 42 seconds. It used Standard discovery, a 90-day window and a two-post per-channel cap. Posts were in Arabic and Persian with an English discovery preference: the preference guides discovery and does not filter posts by language. Three charged discovery pages, five posts and one start give a nominal standard event total of $0.02005 at the checked rates, within a $0.03 run cap. This is one observed example.
For CSV, run a small search, inspect Results and next steps, then export the dataset from Output. Keep the source URL, publication date, channel and text together. Read the summary before broadening the search.
Reuse matching channels without another keyword search
Use Repeat this keyword search in known channels in Output to run the same keyword and filters within the channels that produced matches. Copy the JSON into a new run, review its date range, and set a separate budget. The file carries up to 25 channels and 1,000 delivered post URLs. The next run skips those URLs before applying the result limit and returns an updated input even if it saves no posts. Discovery pages are not needed. A result may be an older post you have not seen. Set a date filter when only recent posts are useful.
After a keyword run saves matches, open Results and next steps in Output. Review the channels and paste the JSON into a new run of this Actor. The file uses up to five channels that produced saved keyword matches, at most one post per channel and five posts total from the last seven days. Channels are selected in collection order, not ranked by relevance; SUMMARY.channelReuse reports omitted channels. If no keyword matches were saved, no input file is available.
This is a fresh channel sample of recent posts from those channels, without the original keywords, date range or phrase/regex filters. Review and adjust it before running; use Telegram New Posts Monitor when you need a checkpoint.
At the current rates and up to 1 GB of memory, five saved posts plus the start event cost $0.00505, with no keyword discovery page fee. Zero posts still incur the start event. Set the new run's maximum cost separately (for example $0.01); the JSON does not carry a spending limit. The original run and optional media in other inputs have their own costs.
Which Telegram Actor should I use?
| Your goal | Recommended Actor | Start with an example |
|---|---|---|
| Find public posts by keyword | Telegram Keyword Search — discovery with literal matches in text or channel title | Product mentions → CSV |
| Export recent posts or accessible history | Telegram Channel Scraper — date filters and CSV, JSON or Markdown layouts | Known channels → CSV |
| Collect newly published posts repeatedly | Telegram New Posts Monitor — reuse the same saved task and checkpoint | New posts → saved checkpoint |
All three work with public Telegram channels, with no Telegram login.
Search Telegram posts in four steps
- Choose What do you want to collect?: Keyword search for topic discovery, Channel scraping for known sources, or Both.
- Fill the matching Keywords or Channels field. A focused mode ignores the other source type and reports a warning if it is filled.
- Set Maximum posts in total. Start with the prefilled five-post Standard search, post-text matching and signature-only exclusion. Existing saved tasks keep their own settings.
- Set Maximum cost per run, start the Actor, and open Output. Choose an export format from Export.
For a keyword-only run that requires the term in the post text and excludes unmatched context posts, use this input:
{"collectionMode": "keyword","keywords": ["bitcoin"],"channels": [],"maxResults": 5,"discoveryCoverage": "quick","keywordMatchScope": "postText","includeContextPosts": false}
The post limit does not cap discovery fees; set Maximum cost per run before scaling up.
Scrape a channel without keyword discovery
{"collectionMode": "channel","keywords": [],"channels": ["cointelegraph"],"maxResults": 20}
When switching a copied or saved task between keywords and channels, check What do you want to collect? at the top of the form. Choose channel-only and clear Keywords for direct channel collection; choose keyword-only and clear Public channels for keyword research. A focused mode ignores the other source type. For API task overrides, include collectionMode, keywords and channels, as in these examples, so the saved mode or unused sources do not carry over. Existing date and phrase filters still apply unless you change them.
Known channels avoid keyword discovery fees. For large history exports, use Telegram Channel Scraper.
When to use it
This Actor is designed for one-time public Telegram data extraction rather than ongoing monitoring. It is useful when you need a current, reproducible dataset for:
- Telegram brand monitoring and competitor research
- News, market, and policy research
- Public-source journalism and OSINT workflows
- Channel archiving for analysis, audit, or compliance research
- Crypto, finance, technology, and community research
- Multilingual topic discovery using country and language preferences
For a repeated, stateful “what is new since the last run?” workflow, use a dedicated monitoring Actor.
Telegram keyword search or channel scraping?
1. Discover public posts by keyword
Add one or more keywords when the relevant channels are not yet known. The Actor finds public Telegram posts and channels related to your keywords and returns posts with a literal match in the post text or public channel title by default. Open Search depth and result limits and choose Quick, Standard, or Thorough coverage to match your time and breadth needs. Country and language preferences help focus discovery ranking; they are not post filters.
Strict matches and optional context
The default keyword output is intentionally strict: every keyword_match row contains at least one requested keyword in its public post text or public channel title. A channel-title match can include posts without the term in their text, and signatures can also match. Matching is literal. Choose Where must the keyword appear? → Post text only in Match quality and text filters for narrower results. Keep Include nearby posts without a keyword off when every discovery result must match the post text.
Enable Include nearby public context posts only when you also need surrounding channel material. Those rows are returned as channel_context, with matchedKeywords: []; they are never represented as literal keyword hits.
2. Collect known public channels
Add a channel username such as cointelegraph, an @username, or a public Telegram channel URL. The Actor collects publicly accessible posts until the date boundary or the run’s global result cap is reached.
3. Combine both in one run
Use keywords and channels together when you need broad discovery alongside authoritative sources. Canonical post URLs are deduplicated, so the same public post is emitted at most once.
How a hybrid run uses the result cap
maxResults is always one global limit. If both sources are present, use Priority when both sources are used to make the allocation explicit:
| Priority | Use it when | Effect if the global cap is reached |
|---|---|---|
| Known channels first | Your selected sources are the most important evidence. | Direct channel collection gets capacity before keyword-discovered channels. |
| Keyword discovery first | Broad topic discovery matters more than source coverage. | Keyword-discovered channels get capacity before direct channel collection. |
You can also set optional per-keyword and per-channel limits to prevent one search term or high-volume channel from using all available capacity. Neither setting changes canonical deduplication.
Real-time Telegram search API for apps and AI agents
Need matching posts inside an HTTP response instead of a dataset? Use the Server tab. Send the same JSON input to POST /search (or GET /search?keyword=openai) and the posts come back in seconds, with the same fields and the same summary as a normal run, ready for brand alerts, AI agents, n8n and Make.
curl -X POST "https://YOUR_SERVER_URL/search" \-H "Authorization: Bearer YOUR_APIFY_TOKEN" \-H "Content-Type: application/json" \-d '{"keywords": ["openai"], "maxResults": 20, "afterDate": "2026-11-01"}'
Copy the server URL from the Server tab. Posts and discovery pages cost the same as in a normal run, and the posts are also saved to the server run's dataset in your account. Normal runs, schedules and dataset exports keep working exactly as before.
The server runs in single-tenant mode: each user gets a separate server run in their own account, and the Server tab shows the active mode. For long keyword lists, media downloads and deep channel history, a normal run is the better fit; if the server is busy, retry after a few seconds.
Input guide
| Input | Use it for |
|---|---|
keywords | Topic, brand, product, organization, event, or phrase discovery across public Telegram content. |
channels | Direct collection from public channel usernames or URLs. |
collectionMode | Guided choice: automatic, keyword-only, channel-only, or hybrid. |
afterDate / beforeDate | Inclusive date boundaries: choose a date or use a relative value such as 7 days. Without any date, the search starts from posts of the last month and then adds older ones. With a window inside the last month, discovery looks for posts of that period first. |
countryCode | Optional ISO 3166-1 alpha-2 preference for keyword discovery. |
languageCode | Optional ISO 639-1 preference for keyword discovery. |
discoveryCoverage | Quick, Standard, or Thorough keyword discovery breadth. Standard is the balanced default. |
maxResults | One global maximum for unique posts from all sources in the run. |
maxResultsPerKeyword | Optional balance limit for verified matching posts from one input keyword. |
sourcePriority | Explicit allocation choice for a combined run that reaches its global cap. |
maxResultsPerChannel | Optional fairness cap for one channel within the global maximum. |
includeContextPosts | Optional, clearly labelled nearby public channel posts; disabled by default. |
includeKeywords / excludeKeywords | Simple phrase filters applied to public post text and public channel title. |
regexFilters | Optional advanced text filters using JavaScript-compatible regular expressions. |
minViews / minSubscribers | Quality filters: keep posts with at least this many public views, or from channels with at least this many public subscribers. |
excludeForwards | Keep only original posts; skip forwards from other channels. |
mediaFilter | Any post (default), posts with media, or text-only posts. |
postLanguages | Keep posts whose text is detected in these ISO 639-1 languages (local estimate, posts of at least 20 characters). |
downloadMedia | Off by default. Enable only when you need available public media files stored with the run. |
More tools by the same author
- YouTube Scraper | Transcripts, Comments, Shorts — follow the same topics on YouTube: videos, Shorts and channels with transcripts and up to 100 comments per video included, from $0.19/1K.
Chinese Simplified
中文说明:Telegram关键词搜索工具(Telegram Keyword Search 爬虫 / API)
Telegram Keyword Search Scraper 按关键词、品牌名或产品名查找提及它们的 Telegram公开帖子,导出帖子正文、发布时间、频道名称和原帖链接,可导出为CSV、JSON或Excel。还不知道该看哪些频道时,用它做品牌提及搜索、舆情监控和话题研究最合适。无需Telegram账号或API密钥。任务名称和输入字段为英文;帖子保留原文,不会自动翻译。
价格透明:每条保存的帖子0.001美元,每个已验证的发现页0.005美元,内存不超过1 GB时每次启动0.00005美元,平台用量已包含。一个已验证搜索页加5条帖子共 0.01005美元。
为什么选择这款Telegram关键词搜索工具?
- 核对提及本身:可选择只在帖子正文中匹配关键词(Post text only),并跳过只出现在已识别频道署名中的匹配,而不是依赖频道标题。
- 来源更多样:设置每个频道的帖子上限,或开启 Spread the sample across channels,避免结果集中在第一个有匹配的频道。
- 费用可控:Maximum cost per run(每次运行最高费用)与帖子数量上限分开设置。
- 重复搜索不再付发现费:在结果指南中打开重复搜索输入,保留原有筛选条件,跳过已交付的帖子,只在已匹配的频道中搜索,不再产生发现页费用。
- 关键词搜索与频道采集结合:可以只搜关键词、只采集已知频道、两者结合,或只在已知频道中搜索关键词;同一条帖子按规范链接去重,只输出一次。
- 匹配依据清晰:每条结果标明
matchType以及在正文或频道标题中匹配到的关键词,方便核实。 - 导出与集成方便:导出CSV、JSON或Excel,或通过Apify API、n8n、Make、Zapier和Apify MCP服务器接入你的系统和AI智能体。
价格对比与实测示例
帖子单价为每1,000条1美元,而 Lofomachines Telegram Keyword Search 标价为每1,000条结果2.50美元,帖子单价低60%;本工具的关键词发现另按发现页计费。以下按2026年11月2日基准测试时核对的标准价格计算(一次启动、内存不超过1 GB):
| 相同数量的保存结果 | 本工具 | Lofomachines(基准价格) |
|---|---|---|
| 已知频道,10条帖子 | 0.01005美元 | 0.02505美元 |
| 关键词,10条帖子,1个发现页 | 0.01505美元 | 0.02505美元 |
| 关键词,5条帖子,1个发现页 | 0.01005美元 | 0.01255美元 |
按这些价格,平均每个计费发现页保存至少4条帖子时,关键词采集更便宜;直接采集已知频道没有发现费。
2026年11月6日实测:主题入门任务在10.4秒内从3个频道保存了5条 bitcoin 相关帖子,费用0.01005美元;随后生成的频道输入又保存了3条近期帖子,费用0.00305美元,没有使用发现页。
适用场景
- Telegram品牌监控和舆情监控:查找提及你的品牌、产品或竞品的公开帖子,逐条打开原帖核实。
- 新闻、市场和政策研究:围绕某个话题收集带日期和来源链接的公开帖子。
- 公开来源调查和OSINT:从关键词出发找到相关公开频道,再用频道采集深入研究。
- 加密货币、金融、科技和社区研究:例如搜索
bitcoin或“比特币”,导出各频道的相关讨论。 - 多语言话题发现:用目标内容使用的语言填写关键词,并可选择国家和搜索语言偏好。
- 频道存档与分析:直接采集已知公开频道,用于分析、审计或合规研究。
可以获取哪些数据?
- 帖子:正文或说明文字、规范原帖链接
messageUrl、消息ID、发布时间、链接、话题标签和提及。 - 频道:频道用户名、公开标题和简介,以及订阅人数、照片、视频、文件和链接等公开计数(
channelInfo)。 - 互动:浏览量的原始显示值(如
4.92K)和换算后的数字viewCount,以及按表情统计的表情回应。 - 媒体:公开媒体引用和预览图链接
previewUrl;需要时可开启媒体下载。 - 匹配信息:
matchedKeywords、matchedTextKeywords、matchedChannelKeywords和matchType(keyword_match关键词匹配、channel_post已知频道帖子、channel_context可选的上下文帖子)。 - 运行报告:
SUMMARY汇总搜索页、匹配和计费情况;Results and next steps 提供可读的结果说明;CHANNEL_SOURCES列出有结果的频道及示例链接;HISTORY_SCRAPER_INPUT可直接用于 Telegram Channel Scraper 继续导出这些频道。
价格
按事件计费,平台用量已包含:
| 计费项目 | 价格 |
|---|---|
| 每条保存的帖子 | 0.001美元(每1,000条1美元) |
| 每个已验证的关键词发现页 | 0.005美元 |
| 每次Actor启动 | 0.00005美元(至少一次,之后每GB内存一次) |
| 示例(内存不超过1 GB,含启动) | 费用 |
|---|---|
| 1个已验证发现页 + 5条帖子 | 0.01005美元 |
| 已验证但没有结果的发现页 | 0.00505美元 |
| 直接采集已知频道3条帖子 | 0.00305美元 |
| 在已匹配频道中重复搜索,保存1条帖子 | 0.00105美元 |
直接采集已知频道不收发现费。帖子数量上限不限制发现页费用,请在 Run options 中设置 Maximum cost per run。请以当前Pricing页面显示的价格为准。
使用方法
- 在 What do you want to collect? 中选择目标:关键词搜索(Keyword search)、已知频道采集(Channel scraping)、两者结合(Both),或只在已知频道中搜索关键词。
- 在 Keywords to find 中填写关键词,每行一个;或在 Known Telegram channels 中填写频道用户名或链接。请使用目标帖子所用语言的关键词,例如搜索中文内容时填写“比特币”。
- 设置 Maximum posts in total,建议先用5条;在 Where must the keyword appear? 中选择 Post text only,让结果正文都包含关键词。
- 设置 Maximum cost per run,点击 Start,完成后先查看 Results and next steps,再从 Export 导出CSV或JSON。
只在帖子正文中匹配关键词的示例输入:
{"collectionMode": "keyword","keywords": ["bitcoin"],"channels": [],"maxResults": 5,"discoveryCoverage": "quick","keywordMatchScope": "postText","includeContextPosts": false}
也可以直接打开现成的示例任务:
- 用中文关键词搜索中文频道:中文关键词搜索任务。示例关键词为“比特币”,搜索语言偏好为中文,最多保存10条帖子;替换为你的品牌、产品或话题即可。
- 查找品牌或产品提及:品牌提及任务。将 NotebookLM 替换为你的品牌名称,最多保存5条帖子,最近90天。
- 研究某个主题:主题搜索任务。将 bitcoin 替换为你的关键词;每增加一个关键词或一种语言的译法,都是一次单独搜索。
- 导出已知频道的近期帖子:频道导出任务。替换频道用户名,收集最近7天内每个频道1条帖子,无需关键词搜索。
中文示例任务: Telegram中文关键词搜索:人工智能频道讨论
countryCode 和 languageCode 是可选的搜索偏好,用于引导关键词发现,不会翻译关键词或帖子;直接采集已知频道不使用这些设置。
常见问题
需要Telegram登录或API密钥吗? 不需要。本工具不会要求填写任何Telegram凭据。
可以同时搜索关键词和采集频道吗?
可以。两种来源可以放在同一次运行中,并按规范帖子链接去重;Collect first when using both sources(sourcePriority)决定达到总上限时优先处理哪种来源。
为什么这条帖子匹配了我的关键词?
matchedTextKeywords 列出在帖子正文中出现的关键词,matchedChannelKeywords 列出在频道标题中出现的关键词。选择 keywordMatchScope: "postText" 后,只保留正文中包含关键词的帖子。
如何在已找到的频道中重复搜索,而不再支付发现费? 关键词运行保存匹配后,在 Output 中打开 Repeat this keyword search in known channels,把JSON粘贴到新的运行中。它保留原关键词和筛选条件,最多包含25个频道和1,000个已交付的帖子链接,跳过已交付的帖子,不需要发现页。
可以用“7 days”代替具体日期吗?
可以。日期字段支持具体日期,或 7 days、2 weeks、1 month、1 year 这样的相对时间。
想持续采集频道的新帖子怎么办? 本工具用于一次性提取。需要按检查点持续采集新帖时,请使用 Telegram新帖监控工具。
可以通过API、自动化工具或AI智能体调用吗?
可以。通过Apify API或Webhook读取默认数据集和 SUMMARY,也可以接入n8n、Make、Zapier;输入、输出和数据集结构说明一致,可通过Apify MCP服务器作为AI智能体的工具使用。
中文说明会翻译采集到的数据吗? 不会。中文说明只是使用文档;采集到的数据保持来源的原始语言(帖子正文按频道发布时的原文呈现)。
其他Telegram工具
- YouTube 视频、字幕与评论采集:在 YouTube 上追踪同样的话题,视频、Shorts 与频道,字幕文本和每个视频最多100条评论已包含,每1,000个视频0.19美元起。
- Telegram频道采集工具(历史导出CSV):已经知道频道时,导出公开频道的帖子和可访问的历史,每1,000条0.20美元,支持CSV、Excel、JSON和Markdown。
- Telegram新帖监控工具:同一任务自动保存检查点,定时采集公开频道的新帖子,可选按主题发现频道。
Chinese Traditional
中文說明(繁體):Telegram關鍵字搜尋工具(Telegram Keyword Search 爬蟲 / API)
Telegram Keyword Search Scraper 依關鍵字、品牌名稱或產品名稱搜尋提及它們的 Telegram公開貼文,匯出貼文內容、發布時間、頻道名稱與原文連結,可匯出為CSV、JSON或Excel。不需要Telegram帳號或API金鑰。
為什麼選擇這款Telegram關鍵字搜尋工具?
- 核對提及本身:可選擇只在貼文內文比對關鍵字(Post text only)。
- 來源更多元:設定每個頻道的貼文上限,或開啟 Spread the sample across channels。
- 費用可控:Maximum cost per run 與貼文數量上限分開設定。
- 比對依據清楚:每筆結果標示
matchType與比對到的關鍵字。
適用場景
- 品牌輿情監測:找出提及你的品牌或產品的公開貼文,逐則開啟原文查核。
- 競品監測:搜尋競品名稱,匯出相關的公開討論。
- 新聞追蹤:圍繞某個話題收集帶日期與來源連結的公開貼文。
- 社群研究:從關鍵字找到相關的公開頻道,再用頻道擷取深入研究。
使用方法
- 在 What do you want to collect? 選擇目標:關鍵字搜尋、已知頻道擷取、兩者結合,或只在已知頻道中搜尋關鍵字。
- 在 Keywords to find 填寫關鍵字,每行一個;請使用目標貼文的語言,例如搜尋中文內容時填寫「比特幣」。
- 設定 Maximum posts in total(建議先用5則),在 Where must the keyword appear? 選擇 Post text only。
- 設定 Maximum cost per run,點擊 Start,先查看 Results and next steps,再從 Export 匯出CSV或JSON。
{"collectionMode": "keyword","keywords": ["bitcoin"],"channels": [],"maxResults": 5,"discoveryCoverage": "quick","keywordMatchScope": "postText","includeContextPosts": false}
現成範例任務:品牌提及、主題搜尋、已知頻道近期貼文、中文關鍵字搜尋。
中文範例任務: Telegram中文关键词搜索:人工智能频道讨论
價格
每則儲存的貼文 0.001美元(每1,000則1美元),每個已驗證的關鍵字發現頁 0.005美元,每次Actor啟動 0.00005美元(至少一次,之後每GB記憶體一次),平台用量已包含。1個已驗證發現頁加5則貼文共 0.01005美元;直接擷取已知頻道不收發現費。
常見問題
需要Telegram登入或API金鑰嗎? 不需要。本工具不會要求填寫任何Telegram憑證。
想持續擷取頻道的新貼文怎麼辦? 本工具用於一次性擷取;需要依檢查點持續擷取時,請使用 Telegram New Posts Monitor。
中文說明會翻譯擷取到的資料嗎?
不會。說明是中文,資料保留來源的原始語言;countryCode 與 languageCode 也不會翻譯關鍵字或貼文。
Country and language preferences
The starter shows No preference for both country and search language. Neither field is required. Empty, omitted or auto values apply no country or language preference.
Choose a country or search-language preference only when it fits your research. These settings can influence discovery ranking; posts are not filtered by location or language, and keywords are searched as written. Direct channel collection does not use these preferences.
Existing Console inputs and saved tasks may retain earlier choices such as US/en. Use No preference to clear that bias explicitly. To test coverage fairly, keep keyword, dates, result caps and country constant while varying only the search-language preference.
The form separates publication dates, optional country/language preferences, search depth and result limits, match quality/text filters, and combined-source/media settings. Maximum posts in total is shared across all sources. The starter's two-post-per-channel cap favors source diversity; clear that optional cap when collecting more matches from one channel. A post limit is not a cost cap: set Maximum cost per run separately.
Example inputs
Public Telegram keyword research
{"collectionMode": "keyword","keywords": ["bitcoin", "crypto"],"channels": [],"afterDate": "14 days","countryCode": "it","languageCode": "it","discoveryCoverage": "standard","maxResultsPerKeyword": 100,"maxResults": 100,"downloadMedia": false}
Scrape known public channels
{"collectionMode": "channel","keywords": [],"channels": ["cointelegraph", "https://t.me/durov"],"afterDate": "7 days","maxResults": 250,"downloadMedia": false}
Combine discovery and direct collection
{"collectionMode": "hybrid","keywords": ["open source security"],"channels": ["cointelegraph"],"sourcePriority": "keywordsFirst","afterDate": "7 days","includeKeywords": ["advisory", "patch"],"excludeKeywords": ["giveaway"],"maxResults": 150}
Telegram data you can export
The default dataset contains one row per canonical public Telegram post. Each record includes source-channel identity, public channel title and description when exposed, canonical channel and post URLs, message ID, publication time when exposed, post text or caption, public counters when reliably available, structured public reactions by emoji and displayed count, media references, links, tags, mentions, matched input keywords, a clear matchType, and collection provenance.
The Actor deliberately distinguishes “not available” from “zero”:
nullmeans the public page did not reliably expose a value.- An empty array means no value was extracted from the public post.
- Textless posts retain
text: null; no author, engagement figure, date, or media type is invented. editedis true only when the public post is marked as edited;authoris the public post signature when the channel shows one;linkPreviewholds the attached link card (URL, site name, title, description) when shown.media[].typecan bephoto,video,document,poll,sticker,voiceorround_video; documents carrytitleandextra, polls carry the question intitle, andmediaTypeslists the distinct kinds. A videodownloadUrlis set when the public post exposes the file.detectedLanguageis an ISO 639-1 estimate of the post text language;isForwardedmirrorsforwardedFrom.
The default key-value store also contains SUMMARY, with discovered channels, examined posts, verified keyword matches, optional context posts, deduplication, and isolated non-sensitive errors.
CLI, API, and MCP-ready contract
The input schema, output schema, dataset schema, and README use the same field names and meanings. This makes the Actor straightforward to configure in the Apify Console, API, CLI, and Apify MCP server. An agent can determine the required source inputs, inspect the expected dataset fields before a run, and retrieve the post dataset and SUMMARY after it finishes.
The Actor itself does not use an AI model. “MCP-ready” here means that external AI agents can discover and call a well-described Apify Actor tool.
Telegram scraper API, n8n, Make and Zapier
Read the default dataset and SUMMARY through the Apify API or webhooks. The standard Actor result format can also be consumed by compatible automation platforms such as n8n, Make, and Zapier, as well as your own JavaScript, TypeScript, Python, spreadsheet, or data-warehouse workflow.
The canonical public post URL makes downstream linking and deduplication straightforward.
Understand discovery work per keyword
SUMMARY.keywordDiagnostics reports each attempted keyword's elapsed collection time, discovery pages read (discoveryPagesRead), saved rows, examined messages, warning codes and individual discovery-page outcomes. Each page includes candidate channels, linked posts, new strict matches and its status. A page status describes the check, not a final charge; use SUMMARY.billing and the Console charge breakdown for billing.
Selection order
Within each channel, newer message IDs are checked first before applying result caps and filters. When Search inside discovered channels is enabled and discovery points to older posts, the Actor also checks the latest matches in that channel, merges unique posts and applies the caps to newer IDs first. Sources follow discovery order rather than a global newest-first order; set a date filter when recency matters.
Responsible use
This Actor collects public Telegram material; keyword discovery finds public Telegram posts that mention your keywords.
Use the Actor in accordance with applicable law, platform rules, and your organization’s policies. Do not use it to infer identities, location, intent, or engagement values that the public source does not expose.
FAQ
Do I need a Telegram login or API key?
No. The Actor does not ask the user for Telegram credentials.
Can I search keywords and scrape channels at the same time?
Yes. Both sources can be included in one stateless run and are deduplicated by canonical post URL.
What does the country or language option do?
They are optional discovery preferences. They help focus keyword discovery; they are not post filters.
Can I use “7 days” instead of a calendar date?
Yes. The date fields accept a calendar date or a relative value such as 7 days, 2 weeks, 1 month, or 1 year.
What happens if keywords and channels both return more posts than the limit?
The Actor returns at most maxResults unique posts. Choose source priority to decide which source is processed first, and optionally add a per-channel cap for a more balanced dataset.
Why is a field null?
The public source did not expose a reliable value, so the Actor reports null instead of an estimate.
Why does matchType matter?
keyword_match is a verified literal match for one or more input keywords. channel_context is an explicitly requested nearby public post and may not contain a keyword. channel_post came from a public channel you entered directly.
Are media files downloaded automatically?
No. The Actor returns public media references when available. Media download is disabled by default.
Does this Actor monitor new posts over time?
It is built for one-time collection. For checkpoints and only-new-post detection, use Telegram New Posts Monitor.
Pricing
The active public plan charges $0.001 per saved post, $0.005 per verified keyword discovery page, plus $0.00005 per Actor start event (minimum one event, then one per GB of memory). Platform usage is included. Check the live pricing panel for the rates applicable to your run.
How keyword discovery is billed
The keyword discovery page fee has been active since October 18, 2026 at 09:00:21 UTC. Always check the live pricing panel for the rates applicable to your run. A confirmed empty search is recorded as evidence and is not charged. Unverified responses and failed requests do not incur this discovery event either. Collecting directly from known channels does not incur a discovery fee.
A per-page discovery log (keyword, page number, verified, channels found, billable) is available in the default key-value store under DISCOVERY; SUMMARY.billing reports verified and charged pages and the billing mode. Discovery is billed only when this event is included in the run's active pricing. Check the live pricing panel before starting. In preview mode no discovery event is charged. A preview total is an estimate, not a platform invoice.
At these rates, a run with one verified discovery page and five saved posts would cost $0.01005 at up to 1 GB of memory. A verified empty search costs $0.00005, the start event only. Additional discovery pages with results add $0.005 each. Actor start is charged once per GB of memory, with a minimum of one event. Set a maximum cost per run to keep spending within your budget; when discovery billing is active, the Actor stops further work when it cannot bill another page or post within that limit.
maxResults limits saved posts. The Actor collects and saves posts after each verified search page, then stops further discovery when the global result cap is reached. A per-keyword cap also stops further pages for that keyword. With channels-first priority, direct channels are collected before keyword discovery and can satisfy the cap without any search. Large keyword lists with few matching posts can still require substantial search work: use focused batches and set a run budget. Keyword metadata reflects the searches completed before a post was saved; a saved post is not charged or emitted again when a later keyword also matches it.
Price examples for known channels and productive searches
The saved-post rate is $1 per 1,000 posts versus the $2.50 per 1,000 results listed by Lofomachines Telegram Keyword Search. That is a 60% lower post rate, not a promise of 60% lower total cost. Our keyword discovery has a separate page fee.
The following calculations use the standard rates checked for our November 2, 2026 benchmark, one start event and no more than 1 GB of memory. Lofomachines figures include its parent Actor's start and result events; they do not assume any additional child-run charge. No additional child-run charges apply. Check the current rates applicable to your account before running.
| Same number of saved results | This Actor | Lofomachines benchmark rates |
|---|---|---|
| Known channel, 10 posts | $0.01005 | $0.02505 |
| Keyword, 10 posts, 1 discovery page | $0.01505 | $0.02505 |
| Keyword, 5 posts, 1 discovery page | $0.01005 | $0.01255 |
At those rates, keyword collection is cheaper when it saves at least four posts per charged discovery page on average (exact threshold: total saved posts must exceed 10/3 times the number of charged pages). Known-channel collection avoids discovery fees. The comparison covers event prices for equal row counts; our default keyword matching is literal in post text or channel title.
Restart behavior
When discovery billing is active and discovery evidence has been saved, a resurrected run stops with DISCOVERY_RESTART_BLOCKED, so discovery is not repeated or charged again. Existing DISCOVERY and dataset records remain available. Start a new run when a new extraction is intended; a new run is separate work with a separate budget.
If discovery billing or evidence storage fails, work stops with DISCOVERY_BILLING_FAILED. Inspect the previous run before starting another: an uncertain billing response is not retried by this Actor.
Run quality and diagnostics
SUMMARY.quality reads bounded when the attempted public scan finished, limited when a selected limit, budget or collection scope bounded the run, and partial when a source error, a known coverage gap or a row without a required post ID, link or timestamp occurred. recordsInspected, missingRequiredFields and optionalFieldsUnavailable are field checks, not billing counters.
A channel whose public content cannot be read is reported as CHANNEL_UNREADABLE, separately from a confirmed empty page. channelScans lists attempted channel contexts, pages read and stop reasons; channelsNotChecked lists direct channel inputs not reached when collection stopped.
Each saved post must belong to the channel being read. Posts from other channels are excluded before keyword matching, result limits and saving, and reported as UNEXPECTED_POST_CHANNEL. SUMMARY.channelPages.sourcePostsIgnored and sourceMismatchPages count them; they are diagnostic counters, not billable events. Forwarded text and links to other channels inside a valid post are preserved.
Temporary source errors are retried automatically. DISCOVERY and SUMMARY.billing report verified pages. A verified discovery page is charged once when discovery billing is active, even if it needed more than one attempt. Pages that could not be verified do not incur the discovery event.
Changes on 8 November 2026
| Before | Now |
|---|---|
UNRECOGNIZED_PUBLIC_PAGE | CHANNEL_UNREADABLE |
| Previous codes for unavailable discovery | DISCOVERY_UNAVAILABLE; DISCOVERY_READ_DEADLINE is still reported in legacyCode |
| Partially checked search results | DISCOVERY_RESULTS_PARTIAL |
channelScans[].stopReasons for an empty channel or the end of in-channel matches | empty_page, search_no_older_posts |
SUMMARY.discoveryRequests | SUMMARY.discoveryPagesRead (also in keywordDiagnostics) |
DISCOVERY page entries | channelsFound; candidateCount keeps the same number |
Detailed technical counters and stage timings were retired from SUMMARY. Use discoveryPagesRead, SUMMARY.billing, DISCOVERY, channelSearch.paginationPostsSaved and performance.elapsedMs / performance.firstSavedResultMs.
How are duplicate keywords and date filters handled?
Repeated keywords, including letter-case variants, trigger one discovery sequence per unique term. Date-only values use UTC: afterDate starts at 00:00:00 and beforeDate ends at 23:59:59.999. Timestamps with an explicit timezone retain that instant. Relative months and years use calendar arithmetic, clamped to the last valid day when necessary. Invalid dates or a start later than the end stop collection with an input error. Undated posts are excluded when a date filter is active.
Why did this post match my keyword?
matchedTextKeywords lists the discovery terms found literally in the public post text or caption. matchedChannelKeywords lists terms found in the public channel title. matchedKeywords contains the terms accepted by your selected scope. These fields are also included in the dataset table and exports.
The default keywordMatchScope: "postText" keeps only posts whose text contains the keyword. Choose "textOrChannel" to also accept channel-title-only matches. For example, a post saying “Market update” in a channel titled “Bitcoin News” matches bitcoin in the default mode, but is excluded in postText mode with context disabled. A signature containing bitcoin still counts as a text match. Phrases must occur within one field, never across the boundary between post text and channel title.
If includeContextPosts is true, unmatched nearby posts may still be included as channel_context with empty matchedKeywords; the evidence columns can still show a title match rejected by postText scope. Direct-channel collection remains unfiltered by this discovery option. Inspect SUMMARY.keywordMatchScope to confirm the selected mode.
{"collectionMode": "keyword","keywords": ["bitcoin"],"channels": [],"keywordMatchScope": "postText","includeContextPosts": false,"discoveryCoverage": "quick","maxResults": 5}
The discovery page fee applies to each verified page, including pages where no posts pass this filter; set a run budget.
Why can a keyword search take longer than a channel export?
Keyword search discovers public sources before reading matching posts, and that discovery step runs even for small post limits. If you already know the channels, choose Channel scraping to collect directly. Use Quick for a smaller discovery scope.
SUMMARY.performance reports elapsed processing time and time to the first confirmed dataset write (null if none), in milliseconds, excluding container startup. It does not affect pricing or result selection.
Real observed output excerpt
Reduced record from an owner verification run observed on 18 October 2026. Field values are retained from that observation and may have changed since then.
{"channelUsername": "bitcoinbullets","messageId": "17316","messageUrl": "https://t.me/bitcoinbullets/17316","publishedAt": "2026-10-08T03:38:47.000Z","views": "4.92K","matchedKeywords": ["bitcoin"],"matchedTextKeywords": ["bitcoin"],"matchedChannelKeywords": ["bitcoin"],"matchType": "keyword_match","discoveredBy": "keyword","scrapedAt": "2026-10-18T14:10:10.306Z"}
Post text and media are omitted here. Both text and channel fields matched this keyword in the observed record; matchedTextKeywords and matchedChannelKeywords show each field separately. The formatted views string is preserved.
Exact phrases and historical links
Literal matching normalizes Unicode, whitespace and common apostrophe variants. Inline highlighting does not insert spaces inside words. Word order and word boundaries matter; plural or reordered forms are separate keywords. Include/exclude phrase filters use the same whitespace and apostrophe normalization; regular expressions still run against the returned post text.
When discovery links to a specific post or an older part of a channel, the Actor checks that part instead of replacing it with the newest posts. A channel-only discovery result checks its newest public posts. Search inside discovered channels (searchWithinChannels) is enabled by default for bounded additional matches while result limits leave room, including channels where linked posts already saved matches: up to three channels per keyword, including an apostrophe variant when needed. This bounded recovery works within result and spend limits. SUMMARY.channelSearch reports whether it was enabled, its per-keyword channel limit and the additional posts saved.
Two consecutive verified discovery pages that add no new channels or posts stop further discovery for that keyword. Pages already verified remain chargeable at the active discovery rate. Use the bounded starter above and an explicit maximum cost for a first test; Quick coverage remains available for a cheaper, narrower search.
For a zero-result run, inspect SUMMARY.resultInterpretation, filterDiagnostics, channelScans, errors and billing. nonMatchingPosts counts examined post occurrences rejected by literal matching; matchingPostsRejectedByFilters counts literal matches rejected by date or phrase/regex filters. They count post occurrences across searches, not unique posts.
Limit discovery pages without new matches
maxUnproductiveDiscoveryPages defaults to 2. After consecutive verified pages without new matching posts, skip remaining discovery for that keyword and continue to the next. Set 0 to disable this match-based limit; duplicate-target and budget guards still apply. Already verified pages remain chargeable. SUMMARY.unproductiveKeywords lists keywords stopped this way.
Validated Telegram destinations
Only validated public Telegram destinations are accepted. Literal keyword matching and all result, exclusion and spend guards always apply. Results that could not be checked within the run's limits are reported in SUMMARY.errors as DISCOVERY_RESULTS_PARTIAL.
Results and next steps
Open Results and next steps in Output for a readable report of the query, filters, checked pages and discovery event counts. Its structured equivalent is SUMMARY.resultGuidance. Zero-result guidance distinguishes source warnings, spend limits and literal matches rejected by filters. Verified discovery pages and the start event are charged even when a run saves zero posts; the run charge breakdown lists them.
The report links to CHANNEL_COLLECTION_INPUT only when saved keyword matches produced that file. Otherwise it explains why no reuse input exists. The report never starts another run. Review any suggested query/filter changes and set a separate budget before retrying.
Review keyword matches in channel signatures
Output separates matchedBodyKeywords, matchedSignatureKeywords and signatureOnlyKeywords. signatureCandidateDetected is a conservative hint: it recognizes a short trailing channel-brand signature after a repeated separator, optionally followed by Trading, Official or News. Original text and default matching stay unchanged.
excludeSignatureOnlyMatches is enabled by default and skips keyword-discovery posts whose text matches occur only in that recognized signature, before consuming result limits. A body match for any queried term keeps the post. This setting does not filter direct channel collection; normal date and other filters still apply. The summary reports filterDiagnostics.signatureOnlyPostsRejected when posts are skipped. Keep it off if a channel-brand mention itself is relevant to your research.
Review sources and reuse the discovery
Use excludeChannels to skip known irrelevant public channels before their posts are fetched, in keyword, channel or combined collection. Usernames, @usernames and public Telegram URLs are accepted. Discovery pages already read remain billable.
CHANNEL_SOURCES lists channels with saved posts, available channel names, saved-post counts, literal matched keywords and up to three example message URLs. Review the examples before choosing sources. Per-keyword saved-match counts are already available in SUMMARY.keywordResultCounts; they count saved matches within the configured keyword allowances.
HISTORY_SCRAPER_INPUT is an input for Telegram Channel Scraper — History & CSV Export, with all returned sources, up to ten posts per channel and 100 total posts. Review sources, dates and the separate run budget before pasting it into that Actor. It starts a new collection of recent posts from those channels, without the search keywords, dates or filters. It is null when no posts were saved.
For a small sample in this Search Actor, the existing CHANNEL_COLLECTION_INPUT remains available. Choose Search to discover public keyword matches, History & CSV to export known channels, and Monitor for repeated collection with its documented checkpoint.
Media preview URLs
media[].url preserves the public source reference and can be a Telegram post link. media[].previewUrl contains the public preview image URL when available, otherwise null. For video media it can be a thumbnail, not a video file. Preview URLs can expire.
Public channel counters
channelInfo includes available subscribers, photos, videos, files and links counters plus the channel description. Numbers expanded from abbreviated values are approximate: 327K becomes 327000, while displayedCounters.subscribers preserves 327K. Missing or unrecognized values remain null. These counters describe the public channel, not the collected sample, and can change between runs.
Numeric post counters
views preserves the public display (for example 4.92K); viewCount expands it to a number (4920). Values expanded from K/M abbreviations are approximate. Reaction count uses the same conversion and retains countText. Missing, ambiguous or unsafe numbers remain null; unavailable counters are never reported as zero.
Custom reactions retain their public customEmojiId as a string and type: "custom". Paid reactions use type: "paid". Their emoji is null when no Unicode glyph is exposed; the Actor does not guess an emoji or resolve custom assets. Standard Unicode reactions keep the existing fields.
channelId preserves the public channel identifier and its sign, as a string. It is not a Bot API chat ID and is not converted into one. It is null when unavailable or not verifiably associated with the post. The same value is also kept as publicPreviewChannelId for existing integrations.
Unavailable posts
An explicit Telegram “Post not found” response is reported as PUBLIC_POST_UNAVAILABLE in SUMMARY, with no dataset row.
Older matches inside discovered channels
When Search inside discovered channels is enabled, a productive channel can continue to older matching posts for the same keyword, within a bounded allowance per run. Collection stops when no new matching post is saved, no older matches exist, or a date, result or spend limit applies.
SUMMARY.channelSearch.paginationPostsSaved reports the older matches saved this way; they are also included in postsSaved.
Default recovery and saved inputs
Internal channel search is enabled when searchWithinChannels is omitted. Set it explicitly to false for potentially lower latency. Existing saved inputs that specify false keep that choice. The starter retains five total posts and at most two per channel; recovery can change which posts fill these limits.
Choose freshness, variety and brand context
- Recent mentions: set Published after to your required date.
- More sources: set Maximum posts per channel to 2 and a larger total limit. This can return fewer posts when only a few channels are found. Clear the channel cap for depth within one source.
- Ambiguous brand names: use Also include at least one phrase, for example coffee / espresso / caffè for Lavazza, and exclude known irrelevant contexts where appropriate. These filters inspect text and channel title and use OR within each list. Multilingual context terms improve matching across languages.
Optional source sampling and hashtag recency
Enable Spread the sample across channels (spreadAcrossChannels: true) to take up to two matches from each of the first five discovered sources before filling the remaining quota. Explicit total, keyword, channel and spending limits still apply; direct-channel collection is unchanged. This option is off by default because it can increase latency. It improves distribution within the discovered sample.
With internal channel search enabled, matches for a hashtag form of your keyword are compared with plain keyword matches before result caps. This can recover newer matching posts that omit the hashtag, within the existing recovery allowance. Set searchWithinChannels: false to avoid this recovery. Set a date filter when you need recent posts overall.
Reuse all matching channels
Open Results and next steps to review sources and the results for each keyword. All groups for repeat keyword search lists executable JSON inputs covering the matching channels, with up to 25 channels in each group. The original repeat-input link points to the first group. Each JSON retains post languages, forward and media filters, minimum views and subscribers, text filters, date settings, and delivered-post exclusions. Relative dates are evaluated again when the next run starts.
Review the channels, then run the selected groups separately. Each run has its own total post limit and spending limit. All groups for recent channel samples provides a separate sample of one recent post per channel, without the original keyword filters. The History Actor input requests up to ten posts per source with an aligned total. No follow-up run starts automatically.