
Plan & post queue
Here you decide what posts go out and when — and approve the ready ones.
1Channel, bot and time
A plan is tied to a channel and a publisher bot and sets the posting time (slot). At that moment an approved post goes to that channel. One project can have several plans for different channels: then the plan cards are collapsed — expand the one you need. If there's a single plan in the project, its card is open right away — there's nothing to hide.
2Time is in the project timezone
The plan time (e.g. 09:00) is interpreted in the project timezone, not the server's. So “09:00” means 9 a.m. in the project's timezone — posts are scheduled and published by it. The timezone is set in “About project” (it defaults from your browser). Below the time field you can see which timezone currently applies; if none is set, time is computed by the server's clock — better to set the timezone explicitly. Times of posts in the queue are shown in that same project timezone.
3Topic and mode per day
For each weekday set a mode: “from sources” (rewrite of found material) or “from the model’s knowledge” (a post on the day’s topic, tied to the date). The day topic is OPTIONAL. If you set it, it narrows selection and writing toward it (for sources it's added to the project topic; for model knowledge it becomes the post topic). If you leave it empty, posts are selected by the project as a whole — by the “Marketing context” (and for model knowledge the marketing context itself is used as the topic). Note: a too-narrow or “wrong” day topic can filter out relevant source material — if posts are scarce, try clearing the day topic and selecting by the project. Days can be turned on and off.
4Auto-fill the plan with the model
To avoid building the plan from scratch, project creation (in the wizard and on site import) offers an “Auto-fill the post plan” option. If it's on and a model is configured, the AI proposes topics and modes for 3–5 weekdays based on your project context — you get a ready draft plan to edit here. You then pick the channel and time.
5🧩 Merging similar drafts into an analysis
The plan option “Merge similar drafts into an analysis” is “depth over stream”. When several drafts on the same topic (from different angles or sources) pile up in the queue, the service combines them into ONE deeper post: it synthesises the material, for a scientific topic checks it against open research bases (prior work, objections, caveats — with links) and notes that the topic is drawing attention. So instead of three-to-five shallow items on one thing you get a single strong analysis.
How it works and what matters:
- 🗂️ ONLY this plan's drafts are touched — approved and scheduled posts are left alone;
- ♻️ the original drafts go to the queue archive (status “Rejected”) with a note — recoverable, nothing is lost;
- ⏱️ at most once an hour and one merge per pass — queue and tokens stay under control;
- 🔔 each merge sends a notification.
Enable it for analytical/scientific channels; pairs well with “Publish posts as an article”.
6🔄 Refreshing stale posts
Drafts get stale over time: they may be left without an image (if image generation once failed), written in the older format (before “articles” and scientific enrichment), and sometimes an autopost “falls” into a draft after a one-off image failure and never goes out again. Once a day the service tidies such posts up on its own:
- 🗓️ it takes a few of the oldest — drafts older than 3 days and approved/scheduled posts whose publish date is already more than a day in the past (calendar-pinned ones are left alone);
- ✍️ rewrites them in the plan's current format (article/plain) with scientific enrichment so they look more solid. Posts you edited by hand are NOT rewritten — your edits are preserved;
- 🖼️ regenerates the image if there wasn't one or the post was rewritten;
- 🚀 recovers a “fallen” autopost (on success re-approves and schedules it) and redistributes overdue posts to future times;
- 🔔 sends a summary notification — check the queue before publishing.
Only a few posts are refreshed per run (to avoid spending extra tokens), so a big backlog is cleared over several days. Ordinary drafts under manual moderation stay drafts — approving them assigns a future date anyway.
7Language and audience region
The post language and audience region are set at the PROJECT level — in «Project info» (not in the plan). They go into the model's context for all the project's posts: text is written in the chosen language and adapted to the region's cultural specifics.
8🌐 Channels in other languages (translations)
In the plan settings you can add one or more channels with a chosen language. Then every ready post is AUTOMATICALLY translated and duplicated into those channels in the selected language:
- 🗓️ time and schedule match the main post; the translation is made from the already-finished text;
- 📅 in the «Posts calendar» such duplicates appear under their channel with a language badge (🌐 EN) so they're easy to tell apart;
- 💡 handy for multilingual projects — you write one plan, and channels in different languages fill themselves.
Each language channel must be connected in «Channels» (with its own publisher bot).
9Holidays
You can enable holiday greetings: the service asks the model whether today is a holiday (national or business-themed) and, if so, prepares a greeting post. The option “skip the main post if a greeting was created” avoids two posts a day.
10Auto-posting (auto-approval)
The “Auto-posting: approve proposed posts automatically” checkbox (below the holiday settings) makes the feed fully autonomous: every generated post is immediately marked “Approved” and published on the plan's schedule — no manual confirmation. Handy when you trust the quality and want the channel to fill itself. Make sure a channel and publisher bot are connected, and remember you remain responsible for published content (AI can make mistakes). Without the checkbox, posts wait for your “Approve” in the queue — the manual-review mode.
11📊 Max posts per day
In the plan you can set “Max posts per day” so the channel isn't flooded when sources are active or auto-posting is on (0 — no separate plan limit):
- ⏱️ Posts are spread evenly across the day — from the plan time to the end of the active window, respecting the project's “quiet hours” (from “Project info”). This applies to auto-posting too.
- 🧱 A minimum interval is kept between posts to one channel (60 minutes by default): even with active sources and trends the channel doesn't “dump” posts in a batch. If the daily limit implies more posts than fit at that interval within the active window, the extras roll over to the coming days.
- 🧭 If there are more suitable materials than the limit, the service picks the MOST RELEVANT and publishes those, skipping the rest.
- 🔔 Project members get a notification “more posts than the daily limit” with a solution: raise the plan limit or narrow the day topic/sources.
The limit works at generation time: no more posts than set are created per day for the plan. A system-wide project publishing limit may also apply.
12What the service will not publish
The point of the service is to find new and useful information, not to fill a channel. A post with nothing to say is worse than no post, so the service deliberately will NOT publish:
- 🚫 posts about nothing. If, after stripping promotion and fluff, no facts, figures or findings are left, the model declines to write rather than inventing text around the headline;
- 🚫 news about the life of an organisation — appointments, awards, anniversaries, event announcements: there is nothing in them for the reader;
- 🚫 filler engagement calls at the end (“What do you think?”, “Drop a comment”, “We'd love your feedback”). A post ends on substance — a fact, a conclusion, or what follows from it. A question is fine only when it is specific and follows from the material;
- 🚫 boilerplate instead of a post. If the model is unavailable, the service does NOT publish template text retelling the headline — it waits and retries on the next pass. The material is not lost.
A side effect: fewer posts may go out than there are items in your sources. That is by design — the feed should consist of things worth reading.
13📰 Publish posts as an article
The “Publish posts as an article” checkbox in the plan turns on Telegram's new format: instead of “a bold first line and a wall of text”, the post goes out as a proper article.
- 🧱 What you get: a real title, subheadings per meaning block, lists, quotes and emphasis on what matters — numbers, terms, conclusions.
- 📝 The post becomes richer, not merely longer: the model is asked to pull more specifics from the source (numbers, dates, methodology, caveats), while every block must add a NEW fact. A short honest post beats a long empty one.
- 📐 The length limit rises from 1024 characters to about 4000 — enough for an article people actually finish. The same limit now applies to translations for language channels: previously a translation was squeezed to 1024, coming out half the length of the original. Telegram allows the full 32,768, but nobody reads that wall.
- 🖼 Images: the generated one goes on top as a poster, and images from the source article can be placed through the text with captions where they genuinely help. If an image adds nothing, the model skips it — empty illustrations are worse than none.
- 👁 In “Plan & queue” such a post already shows its structure — headings, lists and quotes. You edit it as plain text: you edit the source, and the “Preview” button next to it shows how it will look in the channel.
⚠️ The format is new and not supported everywhere. If the channel, bot or Telegram version won't take it, the post STILL goes out — as a regular message. Nothing is lost: formatting isn't worth an unpublished post.
💡 For images to land inside the article body, the service needs a public address (configured by the administrator). Without it the article goes out with text, and the image is attached as usual.
When to enable: for analysis, breakdowns, digests and scientific topics — where there's something to structure. For short news and announcements the regular format fits better.
14🔬 Deep fact-check
The “Deep fact-check” checkbox in the plan raises the reliability of “from sources” posts:
- 📰 the service fetches the FULL article text (not just a short teaser) — the model writes from real facts instead of guessing;
- 🧮 it asks the model to write STRICTLY from the source's facts — no invented statistics, figures or dates; if there's little material, shorter beats inventing;
- ⚠️ if the finished post contains numbers absent from the source, the card shows a “Check the numbers…” warning.
The service also always accounts for the material's publication date (won't present old news as breaking) and, for short teasers, fetches the full article on its own. The trade-off is slightly more tokens and time per post, so enable it for channels where facts are critical.
15Auto-posting: spread evenly through the day
So posts don't go out in a burst (e.g. right after quiet hours, when one sweep produced many at once), the service spreads auto-posts EVENLY across the day: from the plan time (or “now”, if it's already later) until quiet hours begin (otherwise until 22:00), with at least ~20 minutes between posts.
- 📥 When new posts appear, the queue is redistributed again, “compressing” the intervals to fit them into the day; whatever doesn't fit today moves to the following days.
- 🗓️ So a post may go out not exactly at the “plan time” but later the same day — that's the even spread.
- 🚑 Early draining only kicks in for a genuine multi-day backlog (the queue stretches more than 2 days ahead) and works gently.
The manual “Generate drafts” and normal scheduled publishing work as before.
16Formatting: bold, italic, links
Above the post text is a toolbar of formatting icons. Select words and click:
- 𝐁 bold — wraps in **double asterisks** (Telegram-style syntax);
- 𝐼 italic — wraps in *single asterisks*;
- 🔗 link — wraps the selection in text.
The “👁 Preview” button shows how the post will look once published (real bold/italic and clickable links) and toggles back to “✏️ Edit”. You can also type the markup by hand: **bold**, *italic*, text. Both the preview and Telegram (via parse_mode=HTML) turn these into real bold/italic/link, not literal asterisks. The post title is automatically bolded on publish.
17🛡 Source editing (hint)
If, while creating the post, the model spotted self-promotion in the source material (promoting another channel/brand/company) or a debatable, unverified opinion, the card shows an amber “🛡 Source editing” note briefly stating what was noticed and how it was handled (e.g. the ad was removed and the source not linked; a claim was qualified or challenged). This is part of the rewrite rules: the service extracts useful, on-topic facts from the source but does not advertise the source itself and treats unverified opinions critically. If there was nothing of the sort — there's no note.
18Source and image in the card
Under the post text the card shows the source the hook came from. If it's a web link, it's clickable and opens in a new tab (for “from the model's knowledge” or trend-based posts there's a technical marker instead). Hovering over the post image shows the model prompt used to generate it in a tooltip — handy to understand or tweak the visual.
19Post length counter
The top-right corner of a post card shows an “xx of yy” counter: xx is the post's current length in characters (title + text), yy is the recommended length for the selected channel. Per-platform guides:
- Telegram — 1024 (comfortable for a post with an image)
- VK — 2000
- Instagram* — 2200
- Facebook* — 600
- YouTube Shorts — 1000
- X (Twitter) — 280
If the post is longer than recommended, the counter turns amber — a hint to trim (on Telegram a long post with an image goes out as several messages). The value updates live as you edit the text.
20Collapsed details on finished posts
For rejected and already-published posts the card is shown compactly: the full text, the “Why selected” block and the “fallback text” warning are collapsed by default so the queue isn't cluttered with finished posts. Click “Show details” to expand them and “Hide details” to collapse again. Drafts, approved and scheduled posts are shown in full — they still need review.
21Post statuses
A post moves through:
- 📝 Draft — created, awaiting decision.
- ✅ Approved / 🗓️ Scheduled — you confirmed, waiting for its slot.
- 📣 Published — sent to the channel.
Separately:
- ⛔ Rejected — you declined.
- 🔴 Error — Telegram refused the post (e.g. the bot isn't a channel admin).
22🔬 Enrichment from scientific sources
The “Enrich from scientific sources” toggle in the plan deepens posts “from sources” with verified science. For the post's core claim the service searches several open academic bases at once — OpenAlex (250M+ papers), Semantic Scholar (with ready-made tldr summaries), Europe PMC (medicine and biology), Crossref (the DOI registry — confirms a work is real, plus citation counts), and CORE (open-access full texts) when a key is connected — merges the results without duplicates and adds what makes the post stronger:
- 📚 prior and foundational research the topic builds on;
- ⚖️ objections and contradictions — where results were contested or failed to replicate;
- ⚠️ important caveats and limits of applicability;
- 🔗 links to the works themselves (DOI), so the post relies on sources, not guesses.
📎 If the material has a reference list (links like “10.1093/aje/kwt006” or “PMID: 9343002” at the end of the article), the service pulls in EXACTLY the works the article cites — more precise than topic search, giving the primary sources the material rests on.
The model receives these distilled points as extra material and weaves them into the post. Notes:
- best for scientific, analytical and explainer channels; little point for short news;
- better enabled together with “Publish posts as an article” — so objections and links have room;
- the base search is free (no keys); only one short model request per post is spent — a bit more cost;
- 🚫 if a publisher won't serve a scientific article to our server (common — Oxford Academic, Springer, Wiley, etc. block server requests by IP), the service still fetches that work's abstract from an open base by its DOI or title — the post comes out with a scientific basis instead of an “unavailable” note;
- if nothing is found on the topic, the post goes out as usual, without filler.
The model links to primary sources neatly inline — on a keyword (e.g. “a study”); there is no separate “Sources” list at the end of the post. The works used are still saved with the post: the queue card shows a “🔬 Scientific sources” block (to check before approving). The base query is visible in “Model requests” (marked OpenAlex).
Only the post's topic/claim is sent to the open research base — this is ordinary scientific search.
23New post: from a link or manually
The “New post” button in “Plan & post queue” opens two modes:
- ✨ From a link — paste a link to an interesting piece (article, news): the service opens it, takes the gist and, based on it plus the project context (tone, audience, style), generates a ready post and then an image. This is an AI request (spends tokens: text + image); the material is used as a hook and source of facts — the model won't invent details.
- ✍️ Manually — type your own title and text. The “✨ Improve with the model” checkbox: on — the post is first run through the model and adapted to the project context, then the image is generated; off — the text stays as-is and the image is generated right away.
An image is created either way. In both modes you can pick a channel and optionally schedule the publish time (without a time it goes to the queue as a draft).
24Approval and scheduled publishing
Drafts wait for your “Approve” — nothing goes out without confirmation. On approval a post is scheduled for the NEXT plan slot (its time, e.g. 11:00), not “in an hour”; if there are already queued posts, new ones are spread across days, so several approved posts don't go out at the same time. If the project has no plan — a fallback of one hour. The publisher sends approved posts when their time comes; you can also publish right away manually (the publish button). On failure a post gets the “Error” status with a retry option. Re-publishing is protected against duplicates.
25Image together with the post
If a post has an image, it is published to the channel together with the text — a post with a picture and caption. Long text that doesn't fit the caption is sent as a separate message right after the image. The image is sent automatically, no extra setup needed. If the image can't be sent for some reason, the post still goes out — as text, without losing the publication.
26Refine and regenerate
Every post in the queue has “✨ Refine with the model”: type an instruction (e.g. “make it shorter” or “switch to a friendly tone”) and the model rewrites the text while a loader spins, then the result is applied. You can also leave the instruction empty and just click “Refine” — the model then brings the post in line with the project context (tone, audience, style), preserving meaning and facts. Handy for manually-created posts to quickly fit them to the project. Refinement runs with the full project context (topic, tone, audience, taboos, extra context and context files) and in the project language (from “Project info”), preserving the original post's facts — only what you asked for changes. The“🔄 Regenerate” button under the image creates a new picture; you can add a hint for it too. If a post is longer than the channel's recommended length, a “✨” button appears next to the length counter (top-right) — clicking it asks the model to shorten the post to fit, using the project context and preserving meaning and facts; the card is blocked with a spinner during the request. Refinement requires a key for the selected model.
27Refresh from source
When editing a post that has a web link to its source, a «🔄 Refresh from source» button is available. It:
- 🔁 re-crawls the post's source and the links inside it (for “link digests” it pulls in the gist of the linked articles);
- 🧩 gathers up-to-date information and rewrites the post text using the current project context (tone, audience, topics, extra context and files), in the project language;
- 🖼️ leaves the image untouched — there's a separate «🔄 Regenerate» for that.
Handy when the linked material has changed or when the original post came out thin (just link headlines). Posts “from the model's knowledge” or trend-based ones have no button — they have no re-crawlable source. It's an AI request (spends tokens) and needs a key for the selected model.
28Editing an already-published post
A published post can be edited too — and the change is applied to the entry ALREADY PUBLISHED:
- ✏️ on publishing the service remembers the sent reference: on Telegram — the message id (if the post went as a single message), on Strapi — the collection entry id;
- 💾 when you edit such a post and save, the Telegram message's text/caption or the Strapi entry is updated automatically;
- 🖼️ the image of an already-sent post can't be changed — only the text (for long Telegram posts split into several messages, channel editing isn't available).
If the channel/integration doesn't support editing or the reference wasn't saved, the post updates only in the service and the published entry stays as is (with a note).
29Collapsed plan view
Each plan is shown as a collapsed chip: the network icon, the channel and the weekdays. Scheduled days are highlighted, and days when posts actually went out (were published) are marked green. Click the chip to expand it for editing.
30Storage and auto-cleanup of old posts
To keep the dashboard and storage from growing indefinitely, posts that have finished their lifecycle — “Published”, “Rejected” or “Error” — and are older than the retention period (≈30 days by default) are automatically deleted together with their images. This is safe: a published post is already in your channel and doesn't disappear from there. Unfinished posts — “Draft”, “Approved”, “Scheduled” — are never auto-removed. A different period (or disabling cleanup) is configured on the server. The “Recent posts” list and Overview counters reflect posts within the retention window.
* Деятельность компании Meta Platforms Inc., а также принадлежащих ей социальных сетей Facebook и Instagram запрещена на территории Российской Федерации
Didn't find an answer? Write to us — we'll help.
Contact support