Claude Code handles most of the technical work in my tracking projects: it writes sGTM client templates and API-to-server pipelines, and it audits GA4 properties and GTM containers directly through MCP servers. The dev team here is me and a terminal. By now most of my working day happens inside Claude Code: tracking pipelines, ecommerce and payment integrations, email and CRM automations, client websites, internal tools, monitoring, this site included. This post sticks to the tracking side and the projects that made it to production, including the limits I ran into. At the end there is a guest pass so you can test it on your own stack.

What Claude Code is, in tracking terms

Claude Code is Anthropic's agent that runs in your terminal. You describe a task in plain language, and it reads files, writes code, runs commands, and checks its own output until the task is done. The difference from a chat window is where it runs: in your environment, with your files and your API keys. It can run your poller against the real API and keep iterating on the response until the thing works - and you approve what it executes along the way.

The part that matters most for tracking work is MCP (Model Context Protocol). MCP servers connect Claude Code to the tools you already use - GTM, GA4, Stape, Google Ads, Meta. Instead of you pasting screenshots of a container, it reads the actual tags, triggers and variables from the API.

Three tracking projects I shipped with Claude Code

A Stripe purchase tracking template for sGTM

My open-source sGTM client template turns Stripe session IDs into full GA4 purchase events with real session stitching, consent handling, and cookies for 30+ ad platforms. Claude Code wrote most of the sandboxed JavaScript - GTM templates run in a restricted environment where standard JavaScript APIs are unavailable and everything goes through permission-gated APIs you have to require() in. I covered the architecture in Stripe purchase tracking via sGTM.

A YOGO booking API to sGTM pipeline

YOGO Booking released a REST API, and I built a Node.js poller that sends orders, bookings and new customers to an sGTM container in near-real-time. Claude Code handled the boring parts (cursor pagination, rate limits, retries) and wrote the companion client template. Details in the YOGO integration post.

A monitoring layer for sGTM pipelines

Server-side pipelines fail silently: an expired API key or a changed payload, and events stop flowing while every dashboard still looks normal. I built a monitor that watches event flow through an sGTM container and alerts on anomalies, with batching and deduplication so it does not spam. It runs in production for a client today. Claude Code built it across a handful of sessions, tests included.

Tracking audits with MCP servers

This is where the time savings are largest. A container audit used to mean an afternoon of clicking through the GTM interface, exporting GA4 settings, and cross-referencing by hand.

With GTM, GA4 and Stape MCP servers connected, Claude Code pulls the live configuration and checks it in one structured pass: orphaned tags, triggers and variables nothing references, tag consent settings that contradict the CMP, GA4 events that arrive but are never marked as key events, server containers logging PII they should not. I still verify the findings, but the mechanical collection work is gone.

Google has released an official read-only MCP server for Google Analytics, so reporting questions can skip the CSV export step entirely.

Do you need to be a developer?

No. You work with Claude Code in plain language, and it explains what it is doing as it goes. Being able to read code helps when you review its output - and you should review its output - but with the right setup (connected MCP servers, sensible permissions, a few project instructions) I would not call it a requirement. If you can structure a GTM container, you can direct Claude Code.

Where you do need judgment is scoping: it will happily build the wrong thing fast if the brief is vague. The skill is describing what correct looks like, then checking the result against real data.

How to try Claude Code for free

Claude Code is included in the Claude Pro plan ($20/month). If you want to test it before paying, I have three guest passes to give away, and one of them can be yours:

Claim a guest pass - 7 days of Claude Pro, free

Full disclosure: that is a referral link. The pass gives you 7 days of Claude Pro at no cost, and I receive a small usage credit if you decide to subscribe once the trial ends. If the passes are used up, the link will tell you.

FAQ

Can Claude Code set up Google Tag Manager?

Yes, with caveats. Through the GTM API or a GTM MCP server, it can read and modify containers: create tags, triggers and variables, review existing setups, and write sandboxed JavaScript for custom templates. Know the limits, though: the GTM API is heavily rate-limited (25 requests per 100 seconds and 10,000 per day per project), so bulk operations on large containers are slow. And the API does not expose preview mode, so debugging still happens in Tag Assistant. You review and publish the changes yourself.

Can Claude Code audit a GA4 property?

Yes. With a GA4 MCP server connected, it pulls property settings, data streams, events and key events directly from the API and cross-checks them against your GTM container.

Is Claude Code only for developers?

No. It runs in a terminal, but the interaction is plain language. Reading code helps when reviewing output; writing it is not required.

Summary

The sGTM templates, API pipelines and monitoring described above are all in production, and audits that took an afternoon now take under an hour. Sandboxed template JavaScript and production monitoring are things I would not have shipped on my own two years ago; now they are normal projects. If you do tracking work and have not tried Claude Code, the guest pass above is a no-cost way to find out whether it fits your workflow.

Questions about this? Get in touch.