Code Review Graph builds a local dependency graph of your entire codebase for AI coding assistants like Claude Code. It eliminates redundant full-repository scanning and reduces hallucinations by providing precise context. The tool runs entirely locally using Tree-sitter parsing and incremental tracking. It delivers only relevant files to AI assistants via the Model Context Protocol. Developers gain faster, more accurate AI-assisted coding without cloud dependencies.
Code Review Graph solves the context management problem for AI-assisted development. It maps file relationships and imports across JavaScript, Python, Go, and Rust projects. The tool integrates directly with Claude Code and other MCP-compatible assistants. You maintain full privacy as all processing stays on your machine.
Project Repository
Project link:
https://github.com/tirth8205/code-review-graph
How It Works
Code Review Graph uses Tree-sitter to parse source files and extract AST nodes. It builds a dependency graph of file relationships across the repository. Incremental tracking watches for changes and updates only affected parts. The MCP server exposes the graph to AI assistants for context-aware file access.
All processing occurs locally, ensuring no code leaves your system. This local approach eliminates cloud latency and privacy concerns. The tool supports any language with Tree-sitter grammars, covering most mainstream programming languages.
The Verdict
Code Review Graph addresses the scaling problem of AI-assisted coding but requires Tree-sitter language support. It works best for projects with clear import structures. The tool is open-source and freely available on GitHub. For developers using AI desktop assistants, it provides the missing structural intelligence needed for larger codebases.
