Code-review-graph - Local Code Intelligence for MCP

Introduction

In AI-assisted code review, token consumption and context window limits have always been the core bottlenecks holding back efficiency. code-review-graph was built to solve exactly that problem. It is a persistent code knowledge graph tool designed on local-first principles, offering both MCP (Model Context Protocol) and CLI support to deliver efficient, precise local code intelligence for AI code review. By building and maintaining a resident local code relationship graph, the tool compresses average context data to under 8.2x its original size, dramatically cutting token overhead while improving the depth and accuracy of code reviews.

Key Features

• Persistent local code graph: Automatically scans and indexes project code structure, building a call-and-dependency graph across functions, classes, and files, all stored locally.
• 30 MCP tools: Provides 30 dedicated MCP tools covering code search, symbol lookup, dependency analysis, and change impact assessment, seamlessly integrating with any MCP-compatible AI client.
• Smart context compression: When an AI review request comes in, only the code snippets and dependency relationships directly tied to the current change are extracted, achieving an average 8.2x reduction in context size.
• CLI support: A full command-line interface lets developers query and analyze the graph directly from CI/CD pipelines or a local terminal.

Highlights

• Local-first, data secure: All code indexes and graph data stay on your local machine—nothing is uploaded to third-party servers, keeping your code assets fully protected.
• Sharp reduction in token costs: Precision context pruning significantly lowers token usage per AI request, making it ideal for frequent reviews of large codebases.
• Deep code understanding: Instead of relying on shallow file-level context, the tool leverages real call chains and dependency relationships so AI genuinely understands the impact of a change.
• Plug-and-play: Supports major languages and frameworks, requires minimal setup, and integrates quickly with existing AI dev tools like Claude Desktop and VS Code.

Who It's For

This tool is built for heavy AI-assisted development users who frequently rely on AI for code review, refactoring suggestions, or bug detection and want to cut API costs while improving feedback quality. It also serves large project maintenance teams working with hundreds of thousands or even millions of lines of code who need to quickly assess change impact, as well as DevOps and CI/CD engineers looking to embed intelligent review steps into automated pipelines with zero manual intervention via CLI. Finally, it is ideal for security-conscious teams that cannot send code to cloud AI services due to compliance or security policies but still want the benefits of AI-assisted code review.