MediaMarkt MCP
Integrates with MediaMarkt NL's AI shopping assistant, allowing users to ask product-related questions and receive text answers plus structured product information such as product ID, EAN, name, brand, price, currency, delivery time, URL, and image.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
mediamarkt-mcp
MCP server that exposes MediaMarkt NL's "AI-modus" shopping assistant as one tool, ask_mediamarkt.

ChatGPT calling ask_mediamarkt with MEDIAMARKT_UI=widget: the MCP Apps carousel shows the products, the model uses the text answer.
How it works
The MediaMarkt site's chat widget POSTs to https://www.mediamarkt.nl/api/v1/ai-chat (Vercel AI SDK UI-message format, SSE response). No login or cookies are needed, only x-mms-country/-language/-salesline headers and a browser User-Agent. The assistant runs its own tools (search, compare, availability, stores) and streams back MCP-style tool results; this server collects the text answer plus the structured product list (with images).
Unofficial endpoint: it can change or be blocked at any time. See MediaMarkt's AI chatbot terms.
Related MCP server: Periskop MCP Server
Install
npm install
npm run check # live smoke test against mediamarkt.nlRun
Streamable HTTP is the default (stateless, works for ChatGPT, Codex, Claude):
node index.mjs # http://localhost:3000
PORT=8080 node index.mjs
MEDIAMARKT_UI=widget node index.mjs # also serve the product carousel UI (MCP Apps)
node index.mjs --stdio # stdio transport insteadSettings (environment):
Variable | Values | Default | Effect |
| number |
| HTTP port |
| URL | derived from host and port | Public MCP endpoint; separate from the widget origin |
| HTTPS origin | unset | Widget origin; required only when |
|
|
|
|
Connect
codex mcp add mediamarkt --url http://localhost:3000For ChatGPT, expose the HTTP server publicly (e.g. cloudflared tunnel --url http://localhost:3000) and add the URL under Settings → Apps in developer mode. Run with MEDIAMARKT_UI=widget to get the carousel.
Widget mode is opt-in for new users: set MEDIAMARKT_MCP_WIDGET_DOMAIN to the widget host's HTTPS origin. The local LaunchAgent derives both its public MCP URL and widget origin from the machine's current Tailscale hostname while keeping MEDIAMARKT_UI=widget enabled locally.
Tool
ask_mediamarkt({ question, language? }) — language is en (default) or nl; the answer follows the question's language, en also switches product names and URLs to the English storefront. Returns content[0].text (the assistant's answer plus product names, prices and sellers) and structuredContent.products[] (productId, ean, name, brand, price, currency, seller, deliveryTime, url, image).
This server cannot be deployed
Maintenance
Related MCP Connectors
Match shopper intent to a merchant's products and fetch live price, stock and product Q&A.
Routes natural-language shopping queries to merchant storefronts, returns normalized results.
Live product discovery for AI shopping agents, with shopper-approved retailer handoff.
Products, prices, shops and cited sources AI shopping answers recommend, and where your brand ranks.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables natural language queries about Magento store catalog, sales, and customer data through dynamic tools for AI assistants.MIT

Periskop MCP Serverofficial
AlicenseNot gradedqualityDmaintenanceEnables AI agents to perform product discovery from natural language shopping intents, returning ranked products with merchant links without completing checkout.MIT- AlicenseNot gradedqualityBmaintenanceEnables AI agents to ask bol.com's Shophulp shopping assistant product questions in any language, returning assistant text plus structured product recommendations with prices, sellers, ratings, and images. An optional widget mode renders grouped recommendation carousels as an MCP Apps view.36 npmMIT
- AlicenseNot gradedqualityAmaintenanceServes a product store's catalog over MCP tool calls so agents can search products, add items to a cart, inspect the cart, and complete checkout without parsing full HTML pages. This keeps action payloads and token usage far smaller than DOM scraping, especially as catalogs grow, and also works through Chrome's native document.modelContext API.MIT