Do I need to know how to program?
No. The product’s settings are plain text files in the project folder, and your AI agent writes them: you explain in words what you need (“watch this product’s price”, “send me a morning digest”), and it adds the source, the monitor or the rule. The interface asks you to configure nothing — it only shows. Two things are worth doing by hand: start the container following the instructions, and give the agent the management key once. Details are in the Getting Started and Management API sections.
Can I use it without an AI agent?
Yes. News collection, monitors, events, rules, alerts, digests, publishing, feeds and analytics all work without the chat — without it, only the chat itself is hidden. In that case you manage through the API: every command card in the interface can copy a ready command you can run manually, and the settings may be edited directly in the files on disk. Manual editing has a price, though: it is neither validated nor saved to backups, so rework has to be done carefully. Details — Observation Screens and Management API.
Why are there no editing forms in the interface?
This is a deliberate decision, not an unfinished feature. The product assumes the agent maintains the settings while a human observes and commands. That way there are no “half-configured” states: every edit made through the agent is validated, saved to a backup and written to the journal, so it is always visible what changed and when. The interface stays light — you do not have to learn dozens of forms, you simply read and make decisions. Details — Observation Screens.
Will this turn into a stream of notifications?
Noise protection is on from the start; it does not need configuring: at most 3 alerts per hour, a 30-minute pause on a repeat from the same monitor, identical changes — once per day, quiet hours from 23:00 to 08:00 (messages are deferred until morning, not dropped), a separate threshold for bursts. Instead of a stream of messages you can turn on a morning digest. And the main part: every suppressed message is recorded with its reason, so silence is always explainable. Details — Alerts & Noise Protection.
Where does the news come from?
From fourteen source types: RSS, Atom and JSON feeds (the format is recognized by itself), ordinary pages, product and listing pages, services that return data, a saved search query, a sitemap, public Telegram channels, newsletter intake, document files, Reddit, YouTube, podcasts, the Fediverse and review pages. Honest limits: private Telegram channels, X (Twitter), Instagram and TikTok are not supported; for sites behind a login your saved cookies are used — the product does not perform logins itself. Details — Sources.
How does the AI know what matters?
It rates every event on a ten-point scale and sorts it into bands: noise, worth knowing, material, existential. Then your rules take over — what counts as important for you specifically: by sources, topics, words, entities and thresholds. If the AI was not called for an event, an honest “unscored” mark stands there instead of an invented number. The summary is built strictly from the collected material and comes with links to the exact quotes; when there is no backing, an honest note appears instead of text. Details — AI Processing and Rules.
Can the agent publish something on its own?
No. The agent prepares a draft from the material of an event or a story, and the server verifies every quote: invented ones are discarded, and a draft without confirmations is not created at all. Then the draft waits for your approval — and only after it goes to the chosen target: a Telegram channel, an email, an RSS feed, a webhook or a file folder. An honest caveat: unpublishing works differently per target. On Telegram the message is deleted, the file disappears from the folder, the feed rebuilds itself, while an email and a webhook cannot be retracted. Details — Publishing.
Where are the data and settings stored?
Entirely on your server. The database, the archive of collected pages and the published material live in the product’s persistent folder, and the settings are plain text files in the project folder: convenient to keep under version control, and moving to another server is simply copying the folder. Neither an account with us nor someone else’s cloud is needed. Details — Installation and File Configuration.
What about personal data and closed sites?
The processing is done by your AI agent and your model, so the texts of the material go only where you yourself pointed your agent — not to us. For sites behind a login, cookies from your browser are used: the product stores no passwords and performs no logins. In the settings, passwords and tokens are shown masked when read, and the settings backups — where the secrets are real — live in the data folder and are not served through the file manager. Details — Delivery Channels and File Configuration.
What do I need besides the product itself?
A server or a computer with Docker — nothing else is mandatory. The database is embedded, and a separate one is needed only at large volumes. The AI model connects through your Xedant Agent, so if the chat with the agent is not needed, one container is enough. An Xedant Agent license is already included in the product’s license. Details — Installation.
How much does it cost, and what does the license restrict?
The trial period is 30 days free. After that, prices are the same as the neighboring products:
Did not find your answer — take a look at the documentation or start with the installation.