Frequently Asked Questions

Ailyx makes product analytics autonomous. You ship features; the tracking, the reports, and the watching happen without anyone on your team doing them. No instrumentation backlog, no dashboards to build, no analyst in the loop.

Connect a repository. Ailyx reads your code, works out what is worth measuring, and opens a pull request with the tracking written in — reviewed and merged like any other diff. As events arrive it builds the reports worth having: adoption, funnels, retention, failure rates. Then it watches them and tells you when a new feature underperforms or a release starts failing.

Because with those, all of it is still manual. They store and chart the events you send them — deciding what to track, naming it consistently, writing every call, keeping it working as the code moves, then building each dashboard by hand, is your team's job. That was the only option before AI agents. Ailyx does that work itself: it writes the taxonomy and the tracking code, maintains them as your codebase changes, and builds the reports without being asked.

No — connect your GitHub repository and Ailyx audits the events you already have, fixes what is broken, and adds what is missing, in the naming convention you already use. You will get more out of it on Ailyx, though: it is built for AI agents from the ground up, it costs less, and switching is a one-line change. The tracking helper and every event call arrive as a pull request — the only code you write by hand is a single call telling Ailyx who the current user is. You do not have to remove your existing analytics to start.

You install a GitHub app and choose which repositories it can see. It reads that code to work out what is worth measuring, and pushes a branch to open a pull request. It has no access to your running product and never deploys anything.

JavaScript and TypeScript are fully supported: tracking calls are placed at the correct line, scope-checked, and type-checked against your build before a pull request opens. Python and Go are supported with placement and scope checks but no type-check. Other languages are not recommended yet.

Start with your repository — any project on GitHub can use the tracking side, whatever it was built in. For questions about stored data, connect a database you own: Supabase, Neon, Amazon RDS, or any PostgreSQL or MySQL instance you hold the credentials for. Lovable Cloud and Replit Cloud's built-in databases cannot be connected — those platforms do not expose external access, so it is not something Ailyx can add.

An ordinary function call — track('checkout_completed', { plan, amount }) — placed inline where the thing happens, plus one small helper file the first time. It arrives as a pull request you read like any other diff.

Ailyx watches merged pull requests. Calls you removed are retired, calls that moved are re-anchored, and new code gets its own tracking proposal. The taxonomy follows the codebase instead of drifting from it.

Adoption, conversion funnels, retention curves, and failure and empty-result rates, assembled from your events as they arrive. Ailyx then watches them and tells you when a new feature underperforms or a release starts failing.

Yes. Ask in plain English and get a chart back — over your product events, or over a connected PostgreSQL or MySQL database. Refine it by replying in the same conversation.

Yes. Ailyx runs an MCP server, so an assistant can query your product events and your database through it. Setup is in the docs.

Yes. Events arrive as they happen and database queries run live, so you are monitoring the product as it evolves rather than reading yesterday's extract.

Read access to the repositories you select, and permission to push a branch and open a pull request. Nothing else — no access to your running product, no deploys, and no write access to your default branch.

Every proposed change is applied and validated against your build before the pull request opens. Calls that do not place cleanly or do not type-check are dropped rather than shipped, and the pull request says which ones were dropped.

No. Nothing takes effect until you merge and deploy, exactly as with any other contributor's pull request.

No. Your repository is cloned only for the length of an analysis and deleted when it finishes. What persists is what is needed to place and maintain tracking — file paths, event definitions, and the individual lines a tracking call is anchored to — not a copy of your codebase.

Encrypted with AES-256 and decrypted only server-side, at the moment a query runs. They are never exposed to the browser.