An MCP server for Shopify checkout monitoring

The server mirrors the CLI. It offers a safe MONITORING set of tools and an internal OPS set. The monitoring tools are the ones designed to be driven by an assistant:

• list_tests and get_test: see a store's checkout tests, their state and schedule.
• list_runs and get_run: read run history and get the signed video and Playwright trace for a run.
• run_test: trigger a real checkout run on demand.
• job_status: check a queued or running job.

Destructive operations (pause, regenerate, edit a spec) exist too, but are prefixed ops_ so the tier is unmistakable to both a human and a model. They are internal only.

The server speaks MCP over stdio, so you register it the same way as any local MCP server. In Claude Code, add it to your MCP config:

{
  "mcpServers": {
    "revenue-shield": {
      "command": "python",
      "args": ["-m", "rs_cli.mcp_server"],
      "cwd": "/path/to/repo"
    }
  }
}

Claude Desktop supports the same local MCP servers through its custom connector setting. Once it is registered, the monitoring tools appear in the session.

This is a local MCP server you run on your own machine or box, authenticating with your own key. That is an honest and important distinction:

• There is no hosted or remote MCP URL you can point a cloud assistant at yet.
• There is no ChatGPT app, plugin or GPT for this. MCP support today means the Claude family through a locally registered connector.
• It uses an administrative key, so it is an internal and single-operator tool as it stands, not a multi-tenant product.

What you get today is real: an AI assistant that can actually run your checkout tests and show you the proof, not a demo.