There's no public SellerCloud SDK
Every team running on SellerCloud writes the same glue code: an HTTP client, auth handling, pagination, retry-on-401 token refresh, and a handful of helpers wrapped around the REST endpoints they actually use. Then they wire that glue into Claude or whatever AI tool they're using so they can ask questions of it from a chat window.
This package is the assembled, tested version. It exposes seven read-only SellerCloud endpoints as MCP tools so your AI client can call them directly — no glue code, no boilerplate, no maintained-by-one-person internal package.
The seven read-only tools
sellercloud_search_products— query, paginate, filtersellercloud_get_product— full product detail by SKUsellercloud_search_orders— date range, status, querysellercloud_get_order— full order with line itemssellercloud_get_inventory— qty on hand by warehousesellercloud_list_channels— channel registrysellercloud_get_channel_listing— per-channel listing detail
Every tool returns a uniform JSON envelope:
{ "ok": bool, "data": ..., "error": ... } — so your
MCP client gets a predictable contract. Pagination and 401-refresh are
handled transparently. Retries are clamped on 5xx and connection
errors. 44 unit tests cover the auth, retry, and tool-layer behavior.
Distributed as pip install from GitHub at v0.1. PyPI
release once it has soaked.
Who this is for
This MCP server is for you if:
- You run an ecommerce operation on SellerCloud and use Claude (or any MCP-aware AI client) for ad-hoc questions about your catalog, orders, and inventory.
- You're tired of writing the same glue every time and would rather configure once and never think about it again.
- You want a read-only first integration so you can prove value before widening scope.
It's not for you if:
- You don't use SellerCloud.
- You need write operations — those are intentionally out of scope at v0.1.