Every day, the system collects posts from marketplace Telegram channels, filters out ads and duplicates, scores each story’s importance, and sends sellers a short digest for their niche. Tap “Personal AI assessment,” and it explains how a story could affect their business and recommends what to do
MPstats is an analytics platform for marketplace sellers. It wanted to give its users a faster, easier way to keep up with the news: a Telegram agent that reads the channels for them.
Sellers have to keep up with marketplace news: commissions, listing rules, labeling requirements, changes in the law. Most of these updates show up first on Telegram channels, so sellers end up following many of them at once.
The feed turns into noise: ads, the same story reposted across channels, topics that have nothing to do with the seller. A commission change or a new rule is easy to miss, and reading everything eats up time that could go into selling.
The goal was to pull news from the key channels, keep only what matters, boil it down to the essentials, and tailor it to each seller’s niche and business size. Updates had to arrive on a schedule that suits the seller, and the agent itself had to answer the question “How does this affect me?”
Once a day, the system pulls posts from selected Telegram channels. Ads, very short posts, and duplicate stories never make it into the digest.
At signup, sellers pick topics, product categories, and a delivery schedule, and can optionally add their revenue, headcount, taxes, and region. Every day at the time they chose, they get a digest built around their interests, with news sorted by category.
Every story gets an importance score from 1 to 5, a topic, and a one-sentence summary. Critical news is flagged red, everything else yellow. Tap “Personal AI assessment,” and the AI model explains how the story could affect the seller’s business and recommends what to do.
Here’s how one story travels from a Telegram channel to a personalized analysis, step by step.
Overnight, the system goes through the channel list and pulls new posts. Each post gets checked: Does it have enough text to analyze? Is it an ad? Has the same story already run? Similarity is judged by meaning, not wording: the first version stays, and duplicates are flagged and kept out of the digest.
The AI model decides whether the story matters to marketplace sellers, assigns an importance score from 1 to 5, and flags critical items. Then it writes the gist in one sentence and adds a topic and a one-line takeaway. These labels make the feed quick to scan.
Every hour, the system checks whose digest is due, since each seller has their own schedule. News is grouped by category, with critical items in red and the rest in yellow. Tapping a story’s number opens the full version in place of the previous message, and a “Back to news” button returns to the list. Sellers choose the format: everything in one message or each story separately.
How many analyses a seller gets depends on their plan. The 7-day trial includes personalization. The free plan covers general critical news. The paid plan unlocks the personalized digest and up to 5 analyses a day. Limits are counted in requests rather than rubles, so AI costs stay predictable.
Subscriptions are built right into the bot: payments are connected, and the system reminds users when their trial is ending and issues a receipt. We verified the integration with a test payment.
Also used: Celery + Redis — scheduling and queues · CloudPayments — payment processing · Docker Compose — deployment on the client’s server
MPstats now has a product ready to sell as a paid subscription: instead of reading channels by hand, sellers get one digest tailored to their niche.
The same approach works for anyone who needs to track a large stream of news but has no time to read it all.
We build solutions like this around your needs: turnkey AI products, AI agents and automation, integrations, and data.
Our servicesTell us about your project — we’ll talk it through and show you how to make this work for your business
We use cookies and Yandex Metrica to improve our website. Privacy policy
