Mother's missives

What mothermachine is for

November 3, 2026

Most software you pay for every month is ninety percent the same as the thing the next company pays for. A dashboard, a form, a table, some auth, a few integrations. The last ten percent is the part that's actually yours — your fields, your workflow, your weird edge case. And you're renting the whole thing forever to get that ten percent.

That's the problem mothermachine is built around.

People are now building full apps with AI in an afternoon. Some are deterministic — plain code, runs the same every time. Some are AI-native, with a model doing real work inside them. A lot of these end up as public GitHub repos and then just sit there. Nobody can find them, and even when they do, it's hard to tell what the thing actually does or how much work it would take to make it your own.

mothermachine is a directory for those apps. You come here, find something that's close to what you need, pull it down, and configure it. Not build from scratch. Not subscribe. Configure.

What a listing tells you

Every app here is described the same way: what it does, in a controlled vocabulary, what it's built on, and how much effort it takes to make it yours. That last one is the effort badge. Config-only means change a few settings and go. Light-edit means you'll touch some code. Fork-and-modify means it's a solid starting point, not a finished product. We'd rather tell you that up front than have you find out an hour in.

Built for agents too

Here's the part we care about most. The people who build with AI increasingly don't browse — their agent does. So every listing on mothermachine is written to be read by a machine first. There's a manifest for each app, indexes presliced by vertical and by how much is yours to change, and an llms.txt that tells an agent how to query it all in three cheap fetches. An installable skill and an MCP server are next. Point your coding agent at mothermachine and ask "is there anything that already does X?" and it should get a useful answer without burning tokens crawling repos.

The goal is simple. If someone already built most of what you need, you should start from there.

Where this is going

Right now the directory is open-source only, and anyone can list a public repo. Next comes a submit flow so builders can list their own work, and after that a paid tier where repo owners can sell finished apps. That order matters to us. We want the thing to be useful before we ask anyone for money.

If you've built something with AI that other people could use, list it. If you're about to build something, look here first.