Datalyze Tracking Agent

Tracking Agent: a PR check for your analytics tracking

The Datalyze Tracking Agent checks every pull request for what it does to your analytics tracking, and holds the merge until a person signs off. It runs on your own GitHub runner with your own AI model, so your code and data never reach Datalyze. It costs $100 a month, or $125 with the Datalyze MCP.

It reads the change and works out four things: tracking the change needs and doesn't have, existing events the change breaks, events that are now dead, and user properties it affects. Breaking an event is the one it treats as worst, because a silently changed event corrupts every dashboard built on it.

• New behaviour with no event, such as a new checkout step nobody tracks.
• Broken events: an event renamed in place, a call moved so it fires at a different moment, a property whose values changed.
• Dead events and stale user properties: calls the change removed, and profile values nothing sets any more.
• Convention problems: a property that looks like personal data, set where a first-time value needs set_once, an action tracked without its outcome.

Most pull requests need no tracking change. Those pass with a one-line reason, at no model cost when a script can tell, such as a change to styles or docs only.

Nothing reaches your main branch without two decisions from a person: approving the proposal, then signing off on the verified result. The check is a required status check on the pull request, so GitHub holds the merge until both are done.

If verification fails, the agent tries to fix it twice. After that, or when a fix needs a judgement call, it asks your reviewer and offers lettered options. The Tracking Agent docs explain every status and command.

Datalyze sees that your licence key was used, and when. Everything else stays with you.

The method the agent follows is versioned and signed, and the runner checks the signature before using it. You can pin a version, so results never change without you knowing.

On our test suite of 13 seeded repos, it caught 32 of the 33 problems we planted across three runs, and left every clean pull request alone. The planted problems are the ones we find most often in client audits: an event renamed in place, a call that now fires at the wrong moment, a missing entry_point, a dead event, a stale user property, set instead of set_once, personal data in a property, and an action with no outcome event.

It also runs on this site's own repo. The first real pull request it checked replaced an email used as a user id. The agent proposed four changes, one of them breaking, then implemented and verified them once they were approved through the pull request.

$100 a month for a workspace, or $125 a month with the Datalyze MCP. Both renew monthly until you cancel, and cancelling keeps access to the end of the month you paid for. If you already pay for the MCP, upgrading to both mid-month credits the MCP days you haven't used.

A lapsed payment never stops you shipping. The check then passes with a note that the pull request wasn't checked.

If you don't have a tracking plan yet, start with creating a tracking plan or the free event tracking plan generator. For a one-off clean-up of tracking that has drifted, our analytics audit is the faster route.