I spent years building the foundation before I knew what it was for.
WeOS started as an attempt to rethink how business applications work. It became a framework, then a finance app, then an MCP and knowledge graph, and, in the end, a mini-me. This page is the honest version of that path, including the turns.
Business software you could own.
The first version of WeOS wasn't a framework at all. It was an argument: that the common business applications the books, the inventory, the customer records shouldn't be a monthly rental you can never leave. I wanted to rebuild them as open-source apps and distribute them, so a small business could run on software it actually owned.
To build many apps, I needed to make apps easier to build.
A catalog of applications is a lot of applications. So I built a framework to carry the parts they all share: data, history, permissions, APIs, and I named it WeOS too, which I was quite pleased with at the time and have been explaining ever since.
It's written in Go: small binaries, no runtime to babysit, fast enough that a personal instance costs almost nothing to run. That decision is why one developer can stand up a separate database per user without it becoming a second job.
Assistants instead of screens.
When AI became generally available, the catalogue idea changed shape: rather than dozens of applications with dozens of interfaces, each part of a business could have an assistant. Same data underneath, far less software to design.
It was too broad to build all at once, so I picked the part I understood best and went deep instead of wide. Finance.
A finance app on iOS and Android.
WeOS shipped as a mobile app for business finances, integrated with QuickBooks and Square, so it worked against the books and the payments a business already had. That's where the accounting rigor in mini‑me was earned on real ledgers, with real reconciliation, for people whose month-end depended on it.
Three jobs, one framework.
Then I went back to the framework and rebuilt it around what the assistant idea actually needed. WeOS now does three things at once, and mini‑me uses all three.
- MCP server: Claude talks to your instance through it. The conversation half of mini‑me is this layer: your data, your tools, your permissions.
- Knowledge graph: your information stays connected rather than filed into separate boxes, and every change is kept as history, so nothing is quietly overwritten.
- App development framework: the part that makes training a proven pattern into a real interface a day's work instead of a quarter's.
Nothing about it was specific to business.
The foundation didn't care whose ledger it was. Everything I'd built to give a company a private, connected, conversational system worked exactly as well for a household, and a household was the customer I actually had, sitting right there, with too many accounts and no idea what to make for dinner.
mini‑me is WeOS pointed at one person instead of one company.
The engine is open, and it outlives the product.
WeOS is open source under the AGPL. It predates mini‑me by years and it isn't dependent on mini‑me succeeding. If I stopped tomorrow, the engine your instance runs on would still be there, still readable, still buildable by someone else.
That's the whole reason I can say you're never locked in and mean something by it.