Skip to main content

Connecting an AI Tool

Each store has its own MCP page with connection details ready to copy.

Find Your Store's MCP Page​

  1. Open your store in the dashboard
  2. Navigate to Settings > MCP

The page shows your store's MCP endpoint and a tab for each supported AI tool. Copy values from the page rather than typing them. The endpoint can differ by store region, and the page always shows the right one.

Store and Analytics Endpoints​

Every store has two endpoints:

EndpointUse it for
StoreStore data, WP-CLI, performance, change history and connected apps
AnalyticsQuestions about your store's reporting data

They look like this, with your store in place of <your-store>:

https://mcp.urumi.ai/mcp?store=<your-store>
https://mcp.urumi.ai/mcp/analytics?store=<your-store>

Add both to get every tool. Server names include your store, for example urumi-<your-store> and urumi-<your-store>-analytics, so connections for different stores stay separate.

Claude Code​

Run both commands from the Claude Code tab:

claude mcp add urumi-<your-store> --transport http 'https://mcp.urumi.ai/mcp?store=<your-store>'
claude mcp add urumi-<your-store>-analytics --transport http 'https://mcp.urumi.ai/mcp/analytics?store=<your-store>'

Then:

  1. Start Claude Code and type /mcp
  2. Pick the Urumi server and choose to authenticate
  3. Finish the sign-in in your browser
Keep the quotes

The URL contains ?. Without quotes, some shells (including zsh, the default on macOS) fail with "no matches found" before Claude Code runs.

Claude Desktop​

  1. Copy the config block from the Claude Desktop tab
  2. Add it to your claude_desktop_config.json file
  3. Restart Claude Desktop

The block lists both servers:

{
  "mcpServers": {
    "urumi-<your-store>": {
      "type": "http",
      "url": "https://mcp.urumi.ai/mcp?store=<your-store>"
    },
    "urumi-<your-store>-analytics": {
      "type": "http",
      "url": "https://mcp.urumi.ai/mcp/analytics?store=<your-store>"
    }
  }
}

If the file already has an mcpServers section, add the two entries inside it.

Claude.ai​

  1. In Claude.ai, add a custom connector
  2. Paste the Store URL from the Claude.ai tab
  3. Sign in when your browser asks
  4. Repeat with the Analytics URL

Cursor​

Add the config block from the Cursor tab to one of these files:

  • .cursor/mcp.json in your project, for that project only
  • ~/.cursor/mcp.json, for every project

It uses the same mcpServers format as Claude Desktop.

Other MCP Clients​

Any client that supports MCP over HTTP with OAuth sign-in works. Point it at the Endpoint URL from the Other tab, and add the Analytics URL as a second server.

The client finds the sign-in page on its own. The Other tab also shows the discovery URL in case your client asks for it. You do not need to enter any other sign-in settings.

Signing In​

The first time a client connects, your browser opens a Urumi page titled Authorize followed by the client's name:

  1. Sign in with your Urumi account if asked
  2. Review what the client will be able to do
  3. Under Which sites it can access, pick your store
  4. Click Allow

The client gets the access your account has on that store. See MCP Overview.

If something does not work, see MCP Troubleshooting.