Skip to content

MCP Server

The WandStore MCP server connects your store to AI assistants through the Model Context Protocol. Once connected, an assistant like Claude or Cursor gets a tool for every operation in the WandStore public API — campaigns, cohorts, generated experiences, standalone emails, and analytics — and can drive your store in plain conversation.

It is an optional power-user tool. You never need it to use WandStore; the Shopify app covers the full workflow. Reach for it when you want to ask an assistant to pull numbers, draft and launch campaigns, or manage audiences for you.

The hosted server is the simplest path: add one URL and sign in with Shopify. There is no token to copy and nothing to install.

Claude (claude.ai or Claude Desktop) — add a custom connector:

  1. Open Settings → Connectors → Add custom connector.
  2. Name it WandStore and paste the URL:
    https://api.wand-store.com/mcp
  3. Click Connect. A WandStore page opens — enter your store’s myshopify.com domain and sign in with Shopify.
  4. Review what the assistant will be allowed to do and click Allow access. Done — the connector is live.

Claude Code:

Terminal window
claude mcp add --transport http wandstore https://api.wand-store.com/mcp

The first time you use it, Claude opens the same Shopify sign-in + approval page in your browser.

You authenticate by signing into your own Shopify store — exactly like installing an app. WandStore never sees your Shopify password, and the assistant never sees a token. Access is granted per store; an assistant connected to one store can never reach another’s data.

The server exposes one tool per public API operation — 36 tools in total:

Tool groupWhat it covers
store_…Read plan and billing usage, store settings, and analytics rollups for experiences and campaigns.
campaigns_…List, create, update, approve, and launch campaigns; pull every email asset version and the campaign-to-cohort map.
cohorts_…Create and manage audiences — manual lists, Shopify tag cohorts, and AI smart cohorts.
experiences_…List experiences and versions, promote a draft live, roll back to a prior version, or archive.
emails_…Generate, review, promote, and send standalone marketing emails.
meta_…Liveness checks.

Some things to try once connected:

  • “How are my campaigns performing? Summarize the analytics.”
  • “Create a cohort of lapsed VIP customers, then draft a win-back campaign for it.”
  • “Generate a promotional email for this weekend’s sale and show me the draft before sending.”

Every tool declares its behavior to the client: reads are marked read-only, and irreversible operations — launching a campaign, sending an email, deleting a cohort, archiving — are marked destructive, so well-behaved MCP clients ask for your confirmation before running them. Campaign launch also keeps its normal guardrails: a campaign must be prepared and approved before the launch tool will start it.

If you’d rather run the server on your own machine — for scripting, CI, or a client that doesn’t support hosted connectors — install the published package and authenticate with a per-store API token instead of the Shopify sign-in.

  1. Mint a token in Apps → WandStore → Settings → API tokens (shown once — store it safely; it looks like wand_<store>_<secret>).

  2. Register the local server, passing the token as an environment variable:

    Terminal window
    claude mcp add wandstore --env WANDSTORE_API_TOKEN=wand_... -- npx -y @wandstore/mcp

    Or, for Claude Desktop / Cursor / any client with an mcpServers config:

    {
    "mcpServers": {
    "wandstore": {
    "command": "npx",
    "args": ["-y", "@wandstore/mcp"],
    "env": { "WANDSTORE_API_TOKEN": "wand_..." }
    }
    }
    }

The package runs on Node.js 20 or newer and is a single self-contained bundle (the same one the WandStore CLI ships). WANDSTORE_API_URL is an optional override for the API origin; you normally don’t set it.

  • WandStore CLI - The same API surface as terminal commands you can script.
  • Dashboard & Settings - Where to manage API tokens and your store.
  • Campaigns - Plan and launch AI campaigns (also drivable through MCP).
  • Audiences - Build cohorts in the app (also drivable through MCP).
  • Analytics - Understand the performance signals the analytics tools return.