Break down a Wildberries or Ozon niche right in the chat

The agent calls the MPStats API with the token from your personal account: it walks the tree of Wildberries and Ozon categories and subjects, brings up product cards with daily sales, price segmentation, brands, product and category search queries and annual seasonality. On top of that it builds the niche top by revenue, a comparison of your cards against neighbours in the category across two periods and a picture of which price segment the money sits in.

The main thing about MPStats: it is analytics on top of marketplace data, not an export from your own seller account. The upside is that the whole market is visible, competitors whose accounts you have no access to included. The downside is that figures for your own products are presented as a market estimate, and when it comes to your revenue or stock the agent sends you to reconcile with the Wildberries or Ozon account instead of passing them off as accounting data.

There is nothing to write to here: the agent only requests reports and creates or changes nothing in MPStats. The decisions the analysis leads to — raise the price, enter the season, add phrases to the card — are carried out in the Wildberries and Ozon accounts, and the agent states plainly where the edit belongs. Where there is too little category data for a conclusion, it calls that a limitation of the sample rather than presenting it as a trend.

The set of available reports is capped by your plan: once the limit runs out the service answers with code 429 and a Retry-After header, and the agent waits the stated time instead of hammering with requests. A wide analysis — a whole category, dozens of brands — is therefore split into parts, takes longer to assemble than a note on a single card, and the agent says up front how much data it intends to pull. And if the quota is spent, it calls that a limit of the sample instead of passing partial figures off as complete.