Article

Manage Server-Side Tracking from Claude, Cursor or VS Code with MCP

by Berner Setterwall

November 9, 2026

Manage Server-Side Tracking from Claude, Cursor or VS Code with MCP

GDPR Server Tracking now has an MCP server. Connect your AI assistant and set up server-side GTM trackers, check DNS and health, and read traffic stats by asking.

Setting up server-side Google Tag Manager is not hard, but it is fiddly. You create a server container, copy a config string, pick a subdomain, add a DNS record, wait, verify, then go back to your web container and point it at the new endpoint. Weeks later someone asks whether the tracker is still receiving data, and you are back in a dashboard, clicking around.

Today we're launching an MCP server for GDPR Server Tracking. Connect Claude, Cursor, VS Code or any other MCP client to your account, and your assistant can do these steps for you, or tell you exactly what's left to do.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants use tools from other services. Instead of copying information between a chat window and a dashboard, the assistant calls the service directly, with your permission, and works with real data.

Our MCP server lives at:

https://gdpr-server-tracking.eu/mcp

What you can ask

Once connected, you talk to your assistant the way you'd talk to a colleague who knows the platform:

"Set up server-side tracking for shop.example.com." The assistant creates the tracker, then tells you the exact DNS A record to add and what to change in your web GTM container.
"Is the tracker for shop.example.com working?" It checks the setup stage, looks up the DNS record live and reads the traffic from the last 30 minutes.
"Are all my tracking servers healthy?" It goes through every account you can access and flags trackers with wrong DNS or no tracking traffic.
"Which paths got the most requests in the last hour?" It returns request counts, response times, top paths and a per-minute series.
"Move the tracker for shop.example.com to the Small size." It looks up the sizes and prices and makes the change after you confirm.

Health checks that mean something

A tracking server answers health checks from the load balancer all day long. If you count those as traffic, every server looks busy even when no tracking data arrives.

The MCP tools leave health-check requests out of the totals and report them separately. So "healthy" means your tracker is actually receiving tracking traffic, not only that it's switched on. If a server is up but nothing is coming in, the assistant tells you that and points you at your GTM configuration.

Connect in under a minute

Claude: Settings → Connectors → Add custom connector, paste the URL and sign in with Google.

Claude Code:

claude mcp add --transport http gdpr-server-tracking https://gdpr-server-tracking.eu/mcp

Cursor and VS Code: add the URL as a remote (HTTP) MCP server in mcp.json. The setup guide has the exact snippets.

These clients sign you in through a normal OAuth flow: a browser window opens, you sign in with your Google account and choose what the assistant may do. Clients without OAuth support can use a personal access token, created under MCP access in the dashboard.

You decide what the assistant may do

When you connect, you pick between two levels:

Read: the assistant can list your accounts and trackers, check status and DNS, and read stats and credit usage. It can't change anything.
Manage: it can also create, change and delete trackers. These actions affect billing, so the assistant is instructed to confirm with you first, and deleting a tracker requires repeating its host name.

The assistant only reaches the accounts your own user can open, with the same access checks as the dashboard. Access tokens expire after an hour and are renewed automatically, tokens are stored only as hashes, and you can disconnect any app or revoke any token from the dashboard at any time.

What about privacy?

This is a GDPR product, so the question matters. Your visitors' tracking data stays where it always was: on our bare metal servers in Finland. The assistant never sees individual events. It sees what the tools return: tracker settings, setup status, DNS results and aggregated request statistics such as counts, response times and request paths.

What you send to your AI assistant, and where that provider processes it, is governed by your agreement with that provider. If you prefer not to share even aggregated stats, connect with read-only access to a test account, or don't connect at all. Everything still works in the dashboard.