I was working at one of the biggest banks in the world, on an AI team. I thought I'd be building the future. Instead, I spent weeks hunting down people who held tribal knowledge of systems I needed, like a corporate anthropologist with worse coffee. The Swagger was there, technically. Sometimes it had endpoint names. What it never had: any indication of what those parameters actually meant. I needed IDs that the people who built the system didn't even understand ... turned out they were hardcoded in the frontend years ago. So I started using AI to read the codebase. And that's when it clicked: the code has all the business rules, but it's nearly impossible to navigate. Now imagine you have perfect documentation, a perfect Swagger, everything indexed and linked. That's what I started building. The bottleneck was never the model. It was always context. AI just made that embarrassingly obvious. Unlegacy is what came out of that realization. It parses your codebase via AST, enriches it with your docs, and builds a knowledge graph that AI tools can actually use. Better context in, better code out. Would love your feedback and if you've ever spent three weeks understanding a system before writing a single line of code, you know exactly why this exists.

unlegacy

A knowledge graph of your codebase for better AI coding

A knowledge graph of your codebase for better AI coding