Wb MCP Server

What wb-mcp-server does

The wb-mcp-server MCP server connects AI assistants to the Wildberries Seller API. Its 197 tools cover a broad set of seller workflows, including product cards, prices and discounts, promotions, advertising, sales analytics, stock, FBS and DBS orders, click-and-collect orders, warehouse supplies, seller stock, finance, tariffs, reviews, returns, buyer chats, documents, staff users, and connected shops.

The server also includes diagnostic functions. These can inspect token expiry and scopes, check Wildberries hosts, run low-cost probes for API categories, and identify tools that have started failing consistently. That information can help distinguish a local configuration problem from a change or degradation in the Wildberries API.

How it works

Local clients can launch the server over stdio with uvx wb-mcp-server or the installed wb-mcp command. A Docker deployment adds a web dashboard and exposes MCP over SSE at /sse; client messages are sent to /messages. The dashboard provides views for tool calls, errors, response times, shop management, and diagnostics.

Each request can include a shop_id, allowing multiple Wildberries seller accounts to be used in one conversation. With one account, the README says this value can be omitted. Tokens are encrypted with Fernet rather than kept directly in the MCP client configuration.

Setup and configuration

For a local setup, install the package with pip install wb-mcp-server or run it through uvx. Configure the MCP client with the Wildberries API token and a writable DATA_DIR. The token is created in the Wildberries Seller Portal under API access and is described as valid for 180 days.

For the dashboard deployment, clone the repository, copy .env.example to .env, and run docker compose up -d --build. The service listens on port 8001 in the documented example. A health endpoint is available at /api/health, and persistent application data is kept in the Docker wb_data volume. When running from a local checkout without Docker, set DATA_DIR explicitly because the default /data path is intended for the container.

Tools and capabilities

The wb-mcp-server MCP server includes tools for:

• Creating and updating product cards, media, attributes, barcodes, tags, SEO text, and error or blocked-card checks.
• Reading and changing prices, discounts, promotions, advertising campaigns, bids, keywords, balances, and campaign statistics.
• Reviewing sales funnels, product history, stock, regional sales, search queries, and other analytics.
• Managing FBS, DBS, click-and-collect orders, shipments, labels, boxes, warehouse passes, and marking codes.
• Handling reviews, questions, replies, returns, buyer chats, documents, staff invitations, and seller finance data.
• Checking warehouses, intake coefficients, tariffs, storage, connected shops, token status, and API diagnostics.

Limitations and notes

The wb-mcp-server MCP server is focused on Wildberries seller accounts and requires access to the Wildberries Seller API. Client support differs by transport: the README documents direct SSE support for Cursor, Windsurf, and Claude Code, while Claude Desktop requires either a local stdio connection or an mcp-remote bridge for SSE. The client-specific installation guides are currently provided in Russian.

The project documents 197 tools, although other listing material may show a different count. Wildberries API behavior, token validity, and available permissions remain external dependencies. Docker is needed for the documented dashboard deployment; it is not required for the basic stdio installation.