Local Code Intelligence Graph for AI Reviews

code-review-graph is a local-first Python tool that builds a persistent intelligence graph of a codebase for MCP servers and CLI workflows. It helps AI coding assistants inspect only the files, symbols, and relationships relevant to a review or task, reducing context size for large repositories. The project targets code review, repository navigation, and automation, with benchmarks, a GitHub Action, and reproducible evaluation steps.

Devtools code review mcp python developer tools

Stars: 26K
Forks: 2.5K
Issues: 90
PRs: 49

Open on GitHub Archive stats

Local Code Intelligence Graph for AI Reviews

code-review-graph maps a repository into a persistent local graph that AI coding tools can query through MCP or the CLI. Instead of feeding whole files or large folders into a model, it identifies the code relationships most relevant to a task.

Why it matters

Large code reviews often waste tokens on unrelated context. This project focuses on smaller, targeted context windows for review and navigation workflows, especially in larger repositories.

Notable features

- Local-first codebase indexing
- MCP and CLI support
- Persistent repository graph
- GitHub Action support
- Benchmarks for context reduction