code-review-graph

Why we included this project

AI coding assistants are only as useful as the code they can see, and on a large repository they tend to read everything, which burns tokens and slows reviews down. code-review-graph fixes that by parsing your codebase with Tree-sitter and keeping a persistent graph of how functions, classes, and imports connect. It exposes that map to Claude Code, Cursor, Codex, and Copilot over MCP, so when a file changes it can trace every caller, dependent, and test that might be affected and hand the assistant just that slice. Re-indexing is incremental, so a two-file edit on a big project takes a couple of seconds instead of a full rebuild. There is also a one-command installer that detects your installed coding tools and writes the right MCP config for each, plus a GitHub Action for CI. If your team reviews code in large or mixed-language repos, the benchmarked context reductions mean fewer tokens spent and faster feedback.