Code Review Graph

This open source tool, known as CRG, gives AI coding assistants a persistent structural memory of a repository so they stop rereading the whole codebase on every prompt. It parses functions, classes, imports and call chains into a local SQLite graph, keeps it current with incremental reindexing, and serves the relevant slice of context through MCP. During a review, the model receives only the call chains, owners and risks touched by the diff. It is built for engineering teams that rely on assistants such as Claude Code, Cursor, Copilot, Codex, Windsurf and Zed, and one command configures 16 AI coding tools. It supports 39 languages plus Jupyter and Databricks notebooks. CRG computes blast radius, community clusters and risk scores, exposes 30 MCP tools and 5 prompts, and delivers review summaries and impact reports through MCP, a CLI and a VS Code extension with inline alerts. Updates can be triggered by a watch command or git hooks, and merged diffs and reviewer feedback flow back into the graph. It is local first: no code leaves the machine by default. The vendor reports a median context reduction of about 63 times across six open source repositories and a two file reindex of about 2.5 seconds on a 3,000 file repository, measured at pinned commits and described as representative rather than guaranteed. The tool is MIT licensed and installs from PyPI with pip. There is no pricing page and no paid plan is published.