shoppable-mcp

Shoppable MCP Server

Shoppable’s MCP server lets an AI assistant help shoppers find products, build a cart, and hand off a ready-to-buy checkout link, all from inside a chat conversation.

If you are building an AI agent, copilot, or chat experience that wants commerce as a built-in capability, this server gives your assistant the ability to act on a shopper’s behalf without you having to wire up catalogs, sessions, or merchant integrations yourself.

What it is

The Model Context Protocol is an open standard for connecting AI assistants to outside tools and data. Our server speaks that protocol and acts as a bridge between the assistant and Shoppable’s catalog, merchant network, and cart.

When an assistant connects, it picks up a small, focused set of tools. Think of them as verbs the AI can use during a conversation. The tools are intentionally narrow so the model can compose them naturally without needing to understand the plumbing underneath.

What an assistant can do

Product search

Search the live Shoppable catalog by keyword, brand, category, or price range. Results come back ranked, deduplicated by variant, and scoped to the merchants the assistant is configured to shop with.

Typical conversational queries an assistant can answer:

• “Find me a moisturizer under $50.”
• “Show me running shoes from a brand I’ll actually recognize.”
• “What’s in stock at Saks in skincare right now?”
• “I need a birthday gift for a teenager, around $30.”

Merchant lookup

Find a merchant by name or ID, and see what’s in their catalog. Useful when the assistant wants to filter a search to a specific retailer, or sanity-check whether a brand is currently available.

Add to cart

Add a product to a Shoppable cart and get back a checkout URL. The link is pre-populated and ready to share with the shopper. They click, the cart opens with the item already inside, and they finish the purchase in one tap.

Get checkout URL

Retrieve the checkout link for a cart that’s already in progress, in case the assistant needs to surface it again later in the conversation.

Order lookup

Look up the status of a past order if the shopper wants to check in on something they bought.

How a conversation flows

A typical exchange looks like this:

1. The shopper asks for help finding something.
2. The assistant uses the search tool to query Shoppable.
3. Shoppable returns matching products with images, prices, brand, and merchant.
4. The shopper picks one. The assistant calls add to cart and receives a checkout link.
5. The shopper clicks the link, lands on Shoppable’s checkout, and completes the purchase.

The shopper never leaves the chat to browse. The assistant never has to manage sessions or merchant APIs. The whole experience feels native to whatever app the assistant is running in.

A short example

User: Looking for a gift for my mom. She loves skincare. Something nice, under $80.

Assistant: (searches with the query and price filter) Here are a few options I found at Saks and a couple of other partners…

User: The La Mer one looks perfect. Add it to my cart.

Assistant: (adds to cart, receives checkout link) Done. Here’s your checkout link.

User: (clicks, finishes checkout)

What you get back

Search results come back as a structured JSON payload, one entry per product, ready to parse and surface however your app prefers. Each entry includes:

• id and name
• brand and category
• price and salePrice
• merchant
• image (a CDN-hosted URL, safe to embed)
• checkoutUrl (pre-built, ready to share with the shopper)

How that data gets presented is entirely up to you. Render a product grid, drop the items inline in a chat reply, surface a single best match, or feed the results back into the model for follow-up reasoning. The server stays out of your UI’s way.

Types of questions it’s good at

The server is tuned for shopping intents that benefit from being grounded in a real, current catalog. Some examples of the kinds of context a shopper might bring to a conversation that the tools handle well:

• Open-ended product discovery, where the shopper has a vibe but not a brand in mind.
• Constraint-heavy searches that mix budget, category, occasion, or recipient.
• Comparison shopping across a curated set of merchants.
• Gift-finding, where the assistant is helping someone shop for another person.
• Quick “where can I buy this” hand-offs when a product comes up in conversation.

What it isn’t built for: free-form web browsing, payment processing, or anything that would require the assistant to authenticate as a specific shopper. Those parts of the experience stay on Shoppable’s hosted checkout where they belong.

Connecting

The server is live in production at:

https://mcp.shoppable.com/mcp

Authentication is a bearer token issued by Shoppable. To request access, get in touch with your Shoppable contact and let us know a bit about what you’re building.

If you are using Claude Desktop, we have a one-step setup script that wires everything up for you. Ask and we’ll send it over along with a short configuration walkthrough.

Status

The server is live and actively maintained. We add new tools as partners need them, and we’re happy to talk through use cases that don’t quite fit the current set.

For partnership inquiries, integration help, or product feedback, reach out at hello@shoppable.com, or through whoever you usually talk to at Shoppable.