Angehefteter Tweet

built a tool that compresses entire repos for AI agents
rich: 354k tokens → 28k (-92%)
httpx: 64k → 6k (-91%)
requests: 41k → 7k (-83%)
avg 76% reduction. fits in 128k context now.
not truncation, dep graph + AST summaries
uv add codectx
github.com/hey-granth/cod…

English























