

Aary
90 posts

@iimaary
Building in public 🚀 Execution over everything






We cut Claude Code's token usage by 50% with a single command. Two days ago @RaghavChamadiya and I open sourced something we've been building for a while One command and your entire codebase turns into a structured, human readable wiki with confidence scores that degrade as code drifts. Every section knows how fresh it is. If the underlying code changed and the doc didn't, the confidence score drops. You never read stale documentation again. It exposes everything through an MCP server. 8 tools your AI coding assistant can query in real time. Claude Code, Cursor, Copilot. They stop reading raw files and guessing at architecture. They query structured knowledge instead. Early results show ~50% reduction in token consumption because the model stops wasting context on code it already has a map for. Self hosted. Works with local models including Ollama. Your code never leaves your infra. 282 stars in two days. Developers already opening PRs to contribute. If you use AI coding assistants on any non trivial codebase, the context problem is costing you tokens and accuracy every single session. Repowise fixes it at the structural level. GitHub link in replies








first vibecoded billion-dollar company?






















