Wildberries MCP server with plan and confirm
WB Hermes MCP
A 56-tool Wildberries seller MCP server where every change is a two-step plan-then-confirm operation
- 1 ★ on the repository
- Last commit
Broad access
Why this level
- 56 tools cover the seller's prices, orders, ad budget and financial documents
- The two-step confirmation reduces the risk of an accidental change but does not remove it
Install
- Works in
- Services
Manual install
uv sync --all-groups && export WB_API_TOKEN='…' && uv run python -m wb_mcpThe README's local stdio run.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- 56 tools cover the seller's prices, orders, ad budget and financial documents
- The two-step confirmation reduces the risk of an accidental change but does not remove it
This is third-party code. Review the repository files before installing.
What it does
The server is built on wildberries-sdk and offers 56 tools covering the seller's core work: profile, tariffs, product cards, prices, stock and warehouses, FBS orders and supplies, ad campaigns with bids and negative phrases, funnel and reports, balance, financial documents, reviews, questions and chats. Tool names and errors are written in Russian specifically for GLM-class models and the Hermes agent. Every mutating operation is a two-step flow: a wb_plan_* tool only builds a plan, and wb_apply_change applies it once by a confirmation_id kept in process memory. Topping up the ad budget only works through a separate, explicit plan, and wb_describe_operation explains any tool's purpose and a safe example to the model.
Who it is for. For Wildberries sellers who want to automate ads and orders through an agent, but with explicit confirmation of every change.
Good fit when
- You need broad seller functionality in one server
- You want every change to start as a plan rather than an immediate action
- Your agent runs a GLM-class model and needs Russian tool names without SDK jargon
Not a fit when
- You need a packaged PyPI release rather than a source run
- You are not comfortable with confirmation_id living only in one MCP process's memory
Example request
Build a plan to top up the ad campaign budget and show it before applyingMCP
- Transport
- stdio
- Authentication
- API key
Environment variables
WB_API_TOKENrequired, secret- A WB seller token with the right scopes, never passed as a tool argument
Limitations
You need a WB token with the right scopes and Python 3.12. The confirmation_id lives only in process memory, and the plan and apply calls must go through the same live server instance. The project targets the Hermes agent and GLM-class models, though the local run uses plain MCP stdio and works with other clients too. Two xlsx files with a stock export sit in the repo root, which looks like accidentally committed working data rather than part of the code; do not treat them as a template, and be aware of them when cloning.
How to disable. Stop the server process and remove it from your MCP client config, or run hermes mcp remove wb on a Hermes host.
FAQ
Can the agent change a price or top up the budget right away?
No, a wb_plan_* tool builds a plan first, and only a separate wb_apply_change call with the confirmation_id applies it.
Do I need the Hermes agent specifically?
No, the local run uses plain MCP stdio; Hermes is only needed for the ready-made deployment scenario in the README.
Discussion
No questions or reviews yet.
Related
All Wildberries toolsWildberries Seller API MCP server: cards, prices, promotions, ads, orders, supplies, reviews and finance across several stores
Wildberries MCP server (search + execute)
mcp-server-wildberries
Wildberries Seller API MCP server using a search and execute scheme: 235 actions instead of 235 separate tools
Wildberries MCP
wildberries-mcp
MCP server for the Wildberries seller API: products, prices, stock, orders, FBS supplies, analytics, reviews and ads
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
Wildberries MCP server with a CLI and per-section modules
mcp_server_wildberries
A Python MCP server for the Wildberries API with a separate tool module for each account section
OpenAPI-driven Wildberries MCP server
Wildberries MCP Server
A local Wildberries Seller API MCP server built from a 279-method OpenAPI catalog, with calls defaulting to dry-run
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Wildberries MCP server with plan and confirm (this card) | no | API key | 1 | |
| Wildberries MCP server with a CLI and per-section modules | no | API key | 0 | per-section modules |
| OpenAPI-driven Wildberries MCP server | no | API key | 0 | 279 methods |
README in short
The README covers 56 tools across seller sections, the two-step plan-then-apply model with an in-memory confirmation_id, a local run via uv and the WB_API_TOKEN variable, and a detailed Hermes host deployment scenario that pipes the token without saving it to command history and registers the server with hermes mcp add.