Skip to main content
The LoyaltyDog MCP server exposes our REST API as a set of structured tools any Model Context Protocol client can call. It’s an HTTP wrapper around the LoyaltyDog API — no direct database access — and can be pointed at development, staging, or production.

​
What you can do

Once connected, an AI assistant can answer natural-language questions about your loyalty data:
  • “Show me the top ten loyalty programs by active customers this month.”
  • “List gift cards issued in the last 24 hours.”
  • “What’s the wallet-pass install rate for program XYZ?”
  • “Create a new customer record for [email protected] in program ABC.”
The MCP server enforces the same authentication and program-isolation rules as the underlying REST API — it cannot reach data your token does not own.

​
Architecture

┌──────────────┐   MCP   ┌──────────────┐   HTTPS  ┌──────────────────┐
│ AI assistant │ ───────▶│ MCP server   │ ───────▶│ LoyaltyDog API   │
│ (Claude, …)  │         │ (local proc) │          │ api.loyalty.dog  │
└──────────────┘         └──────────────┘          └──────────────────┘
                              │
                              └── bearer token from env
The server runs as a local process started by your AI client. Your bearer token never leaves your machine except in outbound calls to the LoyaltyDog API.

​
Tool surface

The MCP server covers the high-traffic parts of the API:
  • Customers — list, get, create, update
  • Loyalty programs — list, get, update, custom fields
  • Gift cards — issue, redeem, search, balance check, history
  • Wallet passes — issue, list, regenerate, install URL
  • Reporting — generate program reports
For everything else, hit the REST API directly. See the API reference.

​
Next steps

Install the server

Clone, configure, and run the MCP server locally.

Connect a client

Wire up Claude Desktop, Claude Code, Cursor, or Windsurf.