Built inside ilegra as an experiment: pull company knowledge out of the places it was scattered across and let people just ask for it, instead of digging through docs and old channels. Small team, early stage, genuinely exploratory.
Where it started
The idea came out of my own weekend experiment with LlamaIndex. After that the stack settled fast: FastAPI on the backend, React on the front.
The part that mattered more than the stack
I was the architect and project lead, and I barely touched the code. It was close to the first time I worked almost entirely by describing what needed to exist and then reviewing what came back.
That felt strange at the time. Two years later it’s how I work with agents every day, just with people in the loop instead of models. Linus Torvalds has run Linux this way for decades, reviewing and integrating far more than writing. It used to be a rare way to work. It’s becoming the normal shape of the job for anyone who spends their day with AI agents, and I happened to practice it before I knew what it was practice for.