Skip to content
Back to skills

Shopify Mcp

ASecurity

Use Shopify Storefront MCP tools to search catalogs, inspect products, read and mutate carts, fetch policies, and describe shops across any MSApps / OpsAgents client store — multi-tenant, no shop hardcoded. Trigger on: "shopify", "storefront", "myshopify", "cart", "checkout", "products", "catalog", "collection", "shipping policy", "refund policy", "client store", "merchant", or any task that needs live data from a Shopify merchant. Also use when the user wants to prototype a conversational sh...

  • 9 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 7, 2026
ai-agentsgoapibackend

Works with

  • cli
  • api
  • mcp

Security analysis

A100/100

Scanned October 7, 2026

npx -y skills add MSApps-Mobile/claude-plugins --skill shopify-mcp --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Shopify Mcp?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Shopify Mcp
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/msapps-mobile-shopify-mcp/badge)](https://www.skillsdirectory.com/skills/msapps-mobile-shopify-mcp)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: shopify-mcp
description: >
  Use Shopify Storefront MCP tools to search catalogs, inspect products, read
  and mutate carts, fetch policies, and describe shops across any MSApps /
  OpsAgents client store — multi-tenant, no shop hardcoded. Trigger on:
  "shopify", "storefront", "myshopify", "cart", "checkout", "products",
  "catalog", "collection", "shipping policy", "refund policy", "client store",
  "merchant", or any task that needs live data from a Shopify merchant. Also
  use when the user wants to prototype a conversational shopping flow, a
  recommendation agent, or a customer-support bot over a Shopify store.
---

# Shopify MCP — multi-tenant storefront operator

Use this skill whenever Claude needs to act on a Shopify store through the
`opsagent-shopify` MCP server. Prefer these tools over raw `fetch` calls or
the generic `apify-scraping` skill — they are tokened, rate-aware, and
multi-tenant.

## When to trigger

- "What's in `<client-handle>`'s catalog right now?"
- "Find products in shop X under $50 that contain 'linen'."
- "Build me a cart with three of variant ID ..."
- "Read the shipping policy of `foo.myshopify.com`."
- "Describe the storefront of this shop before we swap the theme."
- "Why is `selectModel()` returning undefined on the product page of `<shop>`?"
  → call `shopify_describe_theme` first to confirm which theme + version is
  live before diving into theme JS.

## Mental model

**One install, every store.** Each tool takes an optional `shop` argument of
the form `<handle>.myshopify.com`. If omitted, the server falls back to
`SHOPIFY_DEFAULT_SHOP` (set per-client in the MCP config). Never hardcode a
shop inside the skill — pass it explicitly when the user mentions one.

**Two backends, one surface:**
- Public catalog + cart ops go to Shopify's **Storefront MCP** endpoint
  (`https://{shop}/api/mcp`). No token. Tools: `shopify_search_products`,
  `shopify_get_cart`, `shopify_update_cart`.
- Richer read ops (products by handle, policies, collections, shop metadata)
  go to **Storefront GraphQL**. Needs `SHOPIFY_STOREFRONT_TOKEN`.

If a GraphQL tool errors with "Storefront GraphQL requires a token," surface
that to the user and ask them to paste one into the plugin config — don't
silently degrade.

## Plan → Act → Verify

1. **Plan** — Identify the shop and the intent (read vs cart-mutation vs
   checkout-mutation). Read ops are cheap; cart mutations should be confirmed
   with the user if the cart is real (not a throwaway test cart).
2. **Act** — Call the narrowest tool. Prefer `shopify_search_products` over
   pulling the whole catalog. Prefer `shopify_get_product` for detail pages.
3. **Verify** — For cart mutations, re-fetch with `shopify_get_cart` and show
   the resulting line items. For checkout starts, return the URL and tell the
   user what will happen if they open it.

## Tools at a glance

| Tool | Class | Token? | Purpose |
| --- | --- | --- | --- |
| `shopify_search_products` | read | no  | Search catalog by free text. |
| `shopify_get_product` | read | yes | Product detail by handle. |
| `shopify_get_cart` | read | no  | Cart state by ID. |
| `shopify_update_cart` | cart-mutation | no | Create/add/update/remove lines. |
| `shopify_get_policies` | read | yes | Shipping / refund / privacy / terms. |
| `shopify_list_collections` | read | yes | Category pages. |
| `shopify_describe_theme` | read | yes | Shop + brand + domain metadata. |

## Failure modes to watch

- **No shop configured** → "No shop configured. Pass `shop` argument or set
  SHOPIFY_DEFAULT_SHOP." Ask the user which store, don't guess.
- **Invalid domain format** → reject anything that isn't
  `<handle>.myshopify.com`. Custom vanity domains (e.g. `<brand>.com`) need
  to be resolved to their `myshopify.com` canonical form first.
- **GraphQL 401/403** → Storefront token expired or scoped too narrow.
  Regenerate in Admin → Sales channels → Headless.
- **Rate limits** → Storefront API allows ~60 RPM per shop. If bursting,
  chunk and backoff; never retry blindly.

## Handoff to related skills

- For theme changes (Symmetry install, customizing Liquid), invoke the
  `symmetry-install` skill in this same plugin.
- For Liquid code edits, defer to the `shopify-liquid` reference skill.
- For the app side (OAuth install on a merchant's admin, webhook
  registration), the `opsagent-shopify-mcp` repo README has the CLI flow.

## SOSA notes

- **Supervised** — cart and checkout mutations must be confirmed with the
  user when they touch a real customer cart. Read ops are unsupervised.
- **Secured** — the Storefront token is sensitive and per-shop; never log it
  or include it in tool outputs.
- **Agents** — this skill owns Shopify-facing data ops only. It does NOT
  touch theme files, Liquid, billing, or Partner API — those are other
  skills or the Admin API (not this server).

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.