wb-mcp: Wildberries search and orders through a running Chrome
wb-mcp
An MCP server searches Wildberries products and reads cards, reviews, photos and orders through a tab of an already running Chrome over DevTools
- 0 ★ on the repository
- Last commit
Broad access
Why this level
- The server acts in your signed-in Wildberries session and reads order history and pickup points
- The open Chrome debugging port gives full control of the browser and session and must not be exposed outward
Install
- Works in
- Services
Manual install
git clone https://github.com/oleg909902/wb-mcp.git && cd wb-mcp && npm install && npm run buildThen `npm run stdio`, a connection sample is in `.mcp.json`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The server acts in your signed-in Wildberries session and reads order history and pickup points
- The open Chrome debugging port gives full control of the browser and session and must not be exposed outward
This is third-party code. Review the repository files before installing.
What it does
The server runs requests to Wildberries inside a wildberries.ru tab of a running Chrome through the Chrome DevTools protocol, so it uses the browser session, cookies and anti-bot token. The auth token and `deviceid` are added right in the browser and never leave it. Tools: `search_products` with sorting and filters, `get_search_filters`, `get_product` with prices per size, composition and reviews, `get_reviews`, `get_product_images`, plus `orders_summary`, `active_orders` and `archive_orders`. Orders are read-only, and the parcel pickup code is deliberately never read or returned. Requests run strictly one at a time to trigger the anti-bot less often. On an HTTP 498 answer the server opens a temporary tab, the browser gets a new token and the request repeats. Tool descriptions and replies are in Russian. It runs in stdio mode for Claude Code or over HTTP at `127.0.0.1:3000/mcp` for ChatGPT.
Who it is for. Wildberries buyers and analysts who want to search products, compare prices and track their own orders in a dialog with an AI agent.
Good fit when
- You need products found by query with brand, price and delivery time filters and prices compared
- Product reviews are read 30 at a time sorted by usefulness or rating
- A buyer asks whether orders have arrived and until which day they are kept at the pickup point
- Product photos are needed by the model at a chosen size to review appearance
Not a fit when
- You need the seller cabinet and the official Wildberries API: this works with a buyer browser sign-in
- There is no running Chrome with an open debugging port and a Wildberries sign-in
- You must keep the server on a VPS: the Wildberries anti-bot often challenges datacenter traffic
Example request
Find products for the query "500 ml thermos" with a rating above 4.7 and two-day delivery and compare pricesMCP
- Transport
- stdio, http
- Authentication
- not required
Environment variables
CDP_URL- Chrome debugging port address, http://127.0.0.1:9223 by default
WB_REFRESH_MIN- Refresh the anti-bot token every N idle minutes, 0 disables
Limitations
It needs Node.js 22 or newer and Chrome or Chromium with the `--remote-debugging-port=9222` port and a Wildberries sign-in. The debugging port must never be exposed to the internet: it gives full control of the browser and the WB session. The project relies on internal Wildberries addresses and may break when they change, and the anti-bot sometimes demands a check. Prices and stock refer to the region of the profile main pickup point. No license is stated in the repository. One author, the last commit in September 2026, no stars.
How to disable. Remove the `wb-mcp` server from MCP client settings, stop the process or container and close the Chrome debugging port.
FAQ
How does the server get access to the account?
Through a wildberries.ru tab of a running Chrome. You sign in, and the token and `deviceid` come from `localStorage` inside the browser and never leave it.
What about the order pickup code?
The server deliberately never reads or returns it, since the code can be used to collect parcels. Orders are read-only.
Which environment variables exist?
`CDP_URL` with the debugging port address (`http://127.0.0.1:9223` by default), `HOST`, `PORT` and `WB_REFRESH_MIN` for refreshing the anti-bot token.
How do I run it for ChatGPT?
With `npm start`: the server listens at `http://127.0.0.1:3000/mcp` over Streamable HTTP. Claude Code uses `npm run stdio`.
How does the Docker variant work?
Build the image with `docker build -t wb-mcp .` and run it with `-e CDP_URL=http://<chrome host>:9222`. Chrome must be reachable from the container.
Discussion
No questions or reviews yet.
Related
All Wildberries toolsWildberries MCP search server
Wildberries MCP Server
A lightweight MCP server for searching Wildberries products via its public JSON API: product card, price, reviews with no browser
rf-marketplaces-mcp: Wildberries analytics and a seller cabinet
rf-marketplaces-mcp
An MCP server gives an agent Wildberries data by article without a token and, with a token, seller sales, stock and review replies
Wildberries MCP
wildberries-mcp
MCP server for the Wildberries seller API: products, prices, stock, orders, FBS supplies, analytics, reviews and ads
Wildberries reviews MCP server
WildberriesToolsMCP
An MCP server that pulls Wildberries product reviews by link or SKU, up to 500 reviews in a single JSON response
Read-only Wildberries dashboard MCP
WB Readonly MCP
A read-only Wildberries MCP server: products, ads, orders, stock, finance and reviews, with writes blocked in code, not just prompts
wb-mcp-armine: a read-only Wildberries MCP bridge
WB Readonly MCP
A small Python MCP server gives an agent seven official Wildberries API methods and refuses to work with a token that has write rights
README in short
The README covers capabilities, a list of tools with parameters and filters, Chrome requirements, running over stdio and HTTP, environment variables, the request flow and the code layout.