Skip to main content
Macropay ships a hosted Model Context Protocol (MCP) server, so any MCP-capable assistant or agent can read and act on your account in plain language — “what did we bill last week?”, “create a sandbox product”, “list customers with an active subscription” — without you writing a line of glue code. Because Macropay is the merchant of record, those same agents operate on top of tax-compliant, dispute-managed billing data. The numbers an agent reads back are already net of the sales tax and VAT we remit on your behalf.

​
Why connect over MCP

Conversational ops

Query orders, customers, and revenue from inside your AI client — no dashboard tab-switching.

Agent tooling

Expose Macropay as a first-class tool so autonomous agents can manage products and read billing context.

Safe to experiment

Point at the sandbox server to try destructive actions against test data before touching production.

​
What you’ll need

  • An MCP-capable client — Claude Desktop, Claude Code, Cursor, Windsurf, Codex, or ChatGPT (Developer Mode).
  • A Macropay account. The server authenticates you through your client’s standard MCP login flow on first connect.

​
Choose your server

Macropay exposes two remote endpoints. Pick the one that matches the environment you want the agent to touch.
The production server acts on real customers, orders, and payouts. When you’re prototyping agent behavior, connect to the sandbox endpoint first.
Anywhere a client lets you paste a raw MCP URL, use one of the two above. If a client instead wants a launch command, wrap the URL with mcp-remote:

​
Client setup

Configuration differs slightly per client. Find yours below — each tab shows the production setup, with the sandbox variant alongside it.
Add the server to .cursor/mcp.json:
Once connected, try a read-only prompt first — for example “summarize this month’s orders” — to confirm the agent is reaching the right environment before you let it create or modify anything.