Logimu Shopping MCP README

Logimu Shopping MCP

Give Claude, ChatGPT, and any MCP client real shopping data.

AI assistants are great at comparing products — and terrible at knowing today's price, what's actually in stock, and which "bestseller" is a relabeled generic. This MCP connector fixes that: your assistant gets observed Amazon (US, UK, DE, CA, AU, FR, IT, ES, JP, MX, BR) and Walmart US data — current prices, live stock, real ratings, price/BSR history, and ranked product shortlists — from a continuously refreshed warehouse of 17M+ tracked products.

Free, no signup: 30 queries per day.

Code

https://api.logimu.com/mcp

Quick start

Client	How
Claude (claude.ai / desktop / mobile)	Settings → Connectors → Add custom connector → paste the URL above. Authentication: Sign in now if you have a Logimu plan (every tool, metered on your plan; OAuth, the connector never sees your key), or keep the pre-selected No sign-in for the free tools. OAuth client: keep Use Claude's published identity. Claude can't change this choice later: remove the connector and add it again.
Claude Code	`claude mcp add --transport http logimu https://api.logimu.com/mcp`, then `/mcp` to sign in when a paid tool asks, or `--header "Authorization: Bearer sk_live_…"` to pass the key directly
ChatGPT (developer mode), Cursor, VS Code, anything MCP	Add the same URL as a remote (Streamable HTTP) server.

Then ask things like:

"Find me the best coffee maker under $80" → ranked shortlists grouped by category or price tier
"Is ASIN B0BDHQSZCV a good buy right now?" → price, stock, buy-box seller, 30-day price/stock changes, brand stats
"Well-rated dehumidifiers under $150 with 1,000+ reviews" → structured warehouse filtering

Tools

Tool	Use when	Returns	Lane
`shopping`	"best X", "find me Y under $Z" — you want a shortlist	Ranked, grouped products (category or Budget/Mid/Premium tiers) with real ratings, prices, stock, `observed_at` stamps + brand facets, in ~100ms	Free: 1 of 30 daily
`product`	You have a specific ASIN/item — "good buy?", "price history?", "who sells this?"	Current snapshot + intelligence blocks: 30-day price/stock change events, all-seller offer table, brand stats; full price, rank and buy-box histories with a key (included in the record), or `history_only` for the series alone (0.05 credit)	Free: 1 of 30 daily
`search`	Filtered structured lists — price/rating/review/BSR/FBA filters, sorting	Up to 25 rows on the free lane, across all 11 Amazon marketplaces plus Walmart US	Free: 1 of 30 daily
`serp`	The live Amazon search-results page for a query, right now	Organic grid, sponsored and video blocks, refinements, pagination — a real fetch, seconds	API key: 1 credit per page fetched
`bestsellers`	What sells best in a category right now; Amazon's best-seller category tree	Up to 100 products in Amazon's rank order, the category's breadcrumb name and its subcategories; a root URL or a country alone returns the top-level categories. A real fetch (seconds), cached up to 30 min	API key: 1 credit per page of 50 (top 100 = 2), root list free
`offers`	Who sells a product right now, at what price; the cheapest new or used offer; the buy-box winner this minute	Every offer on Amazon's offers page (every page): seller and seller id, price, shipping, condition, buy box, FBA / FBM / Amazon, Prime, delivery promise and the seller's feedback, by ASIN or barcode. A real fetch (seconds); a no-offers answer is confirmed by a second, independent fetch	API key: 0.5 credit when an offer is listed; no offers, a miss or a failed fetch free
`seller`	A seller id — rating and feedback history, the storefront, the brands it was observed winning the buy box on, connected sellers	Stored profile (0.5) or live (1), +0.5 history; storefront pages 1 each; `brands` / `network` blocks +1	Paid plan
`brand`	Who sells a brand, how contested / Amazon-dominated / FBA-heavy it is, its seller map

The paid-plan tools (seller, brand, category, market_lists, keywords) appear in tools/list only for keys on a paid plan; the stub lists the whole catalog so directories can see it. Every derived figure is an observed count or share from Logimu's own data — never an estimate — and "observed since" is the data horizon, not a founding date. Keyword answers are Amazon's own suggestion ranking as observed — positions within one list, never volume figures and never a product-ranking map.

Full REST API reference (same engine, same data): api.logimu.com/docs

Free lane and paid plans, honestly stated

	Free (no signup)	Paid plan (API key)
Queries	30 / day	Credit-metered — subscriptions from $19/month, one-time credit packs, unconditional 14-day money-back
Data	Warehouse `shopping` / `product` / `search`, cached (fresh, timestamped)	+ `mode=live` on-demand fetches, `serp`, `bestsellers`, `offers`, seller profiles, brand / category intelligence and market lists, batch jobs on the REST API
Signup	None	api.logimu.com

Every record carries an observed_at timestamp — data is observed from our warehouse, not a live retailer feed.
We have no affiliate or seller relationship with anything the tools return. Nothing is ranked because someone paid for it.
Failed or empty requests are never billed, on any tier.

What's in this repo

The live server is hosted — this repo carries the tool catalog, an introspection-only stdio stub (logimu-mcp-stub) so directory build harnesses can enumerate the tools without credentials, and the listing metadata. tools/call on the stub returns a pointer to the hosted endpoint.

Terminal

pip install .
logimu-mcp-stub   # stdio MCP: initialize / tools/list work; tools/call points to the hosted server