OSS
code-review-graph claims benchmarked context reductions and is at 31,577 stars with 133 open issues
tirth8205/code-review-graph (MIT, Python, created 2026-02-26) builds a persistent local map of a codebase and exposes it over MCP and a CLI so coding tools read only the relevant slice, and it took 50 stars today with a push an hour ago. Unlike most repos in this category it claims measured context reductions on reviews and large-repo workflows rather than an unbenchmarked percentage. Given a lab test contradicted a similar headline claim on caveman last week, the benchmark numbers are worth reproducing before citing them.
↳ Follow the thread