Why we built maviapi

maviapi started, like a lot of things, with something breaking at the worst possible moment. There was no grand thesis at first. There was a scraper, a deadline, and a very long night.

The 2am fire

A few years ago I had a small product that leaned on data from a website with no API. The way you do that is you write a scraper, and the way a scraper works is fine until it doesn’t. One night the site shipped a redesign. My selectors pointed at elements that no longer existed, the job started returning empty rows, and because nothing actually errored, nobody noticed until the numbers downstream looked wrong.

So there I was at 2am, diffing HTML in a browser, hand-patching CSS selectors against a page some team had improved without telling me, because why would they. I’d done this exact repair before. I’d do it again. That was the part that started to bother me.

I wasn’t solving a problem. I was paying rent on one, over and over, in the worst currency there is: sleep.

The same fire, everywhere

Once I started looking, the pattern was everywhere. Nearly every developer I knew had a brittle scraper held together with hope somewhere in their stack, and every one of them treated it as a private embarrassment instead of a shared, solvable problem. The web is full of data that has no front door, and the standard fix is for each of us, separately, to keep breaking into the same buildings.

That’s a ridiculous way to spend a generation of engineering time. The maintenance never ends, the knowledge never compounds, and the failure mode is always the same quiet 2am wrong number.

The bet we made

maviapi is a bet that this should be a service, not a chore everyone re-implements. Point us at a site and you get a stable, documented, versioned API: clean JSON, a real contract, and a system whose entire job is to keep that endpoint working when the site changes underneath it. When a page moves, our build agent re-derives the extraction and a person somewhere doesn’t have to wake up.

The deeper bet is about where the work belongs. The fragility of the open web is real and it’s not going away. But it’s the kind of problem that gets cheaper when one team obsesses over it full time and everyone else gets to treat the result as a dial tone. We’d rather be that team than have you be it at 2am.

Where we are now

Today maviapi runs a growing catalog of hosted APIs, from Reddit and Hacker News to live ship positions off MarineTraffic, and an agent that can stand up a new one on request. None of it is finished. The web keeps changing, which is precisely the point: that change used to be your problem, and now it’s ours.

If you’ve ever patched a selector at 2am, you already understand exactly why this exists. We built the thing we wished we’d had that night.