Break down Wildberries sales and stock right in chat
Connect Wildberries to the AI agent: products, orders and analytics. The agent finds where revenue dropped, breaks down returns and computes unit economics.
How the agent works with Wildberries
The agent connects to Wildberries with a single API token and covers the whole seller loop: FBS and FBW stock, the sales report, unit economics, advertising, reviews, questions, chats, prices and shipments. For stock it pulls a summary across every seller warehouse and FBW balances per article number in one call rather than looping over warehouses, and builds a replenishment plan on top of it that accounts for sales velocity and safety stock. WB spreads those sections across separate APIs — content, statistics, analytics, prices, reviews — and the agent works out which one to ask instead of leaving that for you to remember.
Money is counted from the v5 sales report with correct cursor pagination: WB returns a field that looks like a cursor, but the real cursor is the identifier of the last row — the agent knows this, so it neither loops forever nor drops rows. The report is served no more than once a minute, so long periods are assembled in chunks with the agent holding the pauses itself instead of running into 429s. Commission, logistics and storage are taken from that same report rather than from a price list, so SKU margin is computed on the amounts actually withheld.
Unit economics per article number is assembled with the real advertising spend. On WB advertising lives in two independent loops — standard campaigns in advert-api and media advertising in advert-media-api — and the agent queries both, otherwise half of the spend would simply be missing from the report. On top of that it computes ad-cost share by SKU with alert levels and a month-by-month P&L for the account.
Reviews, questions and buyer chats are collected into a single feed of unanswered items: the agent replies to a review and to a question, pins the best review on the listing and measures your response speed. The reviews section on WB is limited to one request per second, so a long feed is worked through in portions; the agent learns about a new review or question from a webhook rather than by polling in circles. The unanswered feed can be put on a schedule so the summary arrives on its own.
The most valuable part is that the agent knows the character of the WB API. WB can answer «success» to a request with a typo and silently do nothing, so after every write the agent re-reads the state and confirms the change was applied. A listing update is a full overwrite, and the agent never sends a partial one; a media upload fails entirely because of a single broken URL, so links are checked in advance. While an asynchronous card upload is still running (up to 30 minutes), the agent leaves prices and stock alone — otherwise changes are applied only partially. Hard limits such as one FBS pass per 10 minutes and one warehouse rebinding per day are factored into its planning. And anything that changes data — prices, bids, replies — is first shown as a plan and applied only after your explicit confirmation.
Integration scenarios
Find where revenue dropped on WB
The agent compares sales week over week, breaks the drop down by category and SKU and separates seasonality from problems with the listing or stock.
Returns broken down by reason
The agent collects returns for the period, groups them by reason and article number and shows which listings are generating losses.
Review and rating monitoring
The agent checks new reviews, drafts replies and warns you when a listing rating starts sliding.
Unit economics by article number
The agent brings WB sales, commissions, logistics and storage into a single table and computes the real margin on every SKU.
Replenishment plan for the warehouses
The agent pulls FBS and FBW stock across every warehouse, matches it against sales velocity and shows what to ship and in what volume before you run out.
Where the ad budget goes
The agent adds up standard and media campaign spend, computes the ad-cost share per article number and flags where advertising eats more margin than it should.
Wildberries examples
Девять артикулов уйдут в ноль за неделю, ещё шесть — за две.
- Сводка по всем складам FBS и остатки FBW собраны одним вызовом, а не обходом складов по очереди
- Считал по скорости продаж за 14 дней со страховым запасом: «Худи оверсайз» держится 4 дня при плече поставки 9
- План пополнения на 2 340 единиц собрал черновиком — отправлять не стал
How to connect
Before you connect
- An active seller account at seller.wildberries.ru.
- A decision on the agent's job: analytics only, or reviews and prices too.
- The token is stored encrypted and can be revoked in WB at any moment.
- 1
Sign in to WB Partners as the owner
Open seller.wildberries.ru and sign in as the account owner: only the owner issues and deletes API tokens.
- 2
Open “Settings → API access”
Hover over the seller name in the top-right corner and go to “Settings → API access”. The page lists issued tokens and has the “Create new token” button.
- 3
Create a token of the “Service” type
Name the token, pick the “Service” type and find our service in the WB solutions catalogue — categories fill in for you.
- 4
If you picked “Basic” — tick the categories
The “Basic” type needs categories ticked by hand: Content, Statistics, Analytics, Prices and discounts, Questions and reviews.
FAQ
An API token from the seller account: Settings, then API access, issued by the account owner. Read permissions for the Statistics, Analytics and Content categories are enough for analytics. Choose the Service token type and find our service in the solutions catalogue — the categories will be filled in for you: for the Basic type WB cut request limits on 30.03.2026, and long reports then take weeks to assemble.
No. WB issues tokens with granular permissions per category — you choose which sections to open. The token is stored encrypted and can be revoked in the WB account at any moment.
Sales and stock analysis, returns breakdowns, review monitoring, supply reconciliation and unit economics that account for commissions and logistics. The result is a table, a report or a regular summary in Telegram.
The agent queries the API on every task, so it always works with current data. Regular reports can be put on a schedule and the agent will assemble and send the summary itself.