tirth8205/code-review-graph is a local-first code intelligence graph that builds a persistent map of codebases for efficient AI coding tool integration and review workflows.
Source: Official README and linked documentation View on GitHub →This project is being discussed because its focus on reducing token waste in AI coding tools, addressing the pain point of inefficient code reviews in large repositories. It differs through its use of Tree-sitter for code parsing and the Model Context Protocol (MCP) for precise context delivery, offering a unique solution for AI-assisted code reviews.
Source: Official README and linked documentationBuilds a structural map of the codebase using Tree-sitter, tracking changes incrementally and providing precise context to AI assistants.
Source: Official README and linked documentationTraces the 'blast radius' of code changes, identifying all callers, dependents, and tests affected, allowing AI assistants to read only relevant files.
Source: Official README and linked documentationEnables incremental updates in seconds by re-parsing only files with changed SHA-256 hashes, optimizing performance for large codebases.
Source: Official README and linked documentationThe architecture involves parsing the repository into an Abstract Syntax Tree (AST) with Tree-sitter, storing it as a graph of nodes and edges, and querying the graph at review time to compute the minimal set of files needed. It uses the Model Context Protocol (MCP) for context delivery and supports various AI coding tools through platform-native hooks and skills.
Source: Official repository tree and dependency filesmcpfastmcptree-sittertree-sitter-language-packpyyamlnetworkxwatchdogThis project is for developers working on large codebases who need efficient AI-assisted code reviews. It is useful in scenarios where AI coding tools are used for code analysis, review, and refactoring, and where reducing token waste and improving review efficiency is critical.
Latest v2.3.9 was released on 2026-09-18. This edition adds local graph visualization, path and depth boundary, forget command, single platform uninstallation, Voyage embedding provider, and confidence interpretation for the empty results of the three main query tools; the graph schema, client configuration, and existing indexes still need to be checked when upgrading.
Source: github.com/tirth8205/code-review-graph/releases/tag/v2.3.9status: docs-only | checked_at: 2026-09-23 | environment: Read the tirth8205/code-review-graph official README, v2.3.9 Releases, Usage, FAQ, GitHub Action, VS Code and benchmark instructions; package not installed, local code base not scanned. | limitations: No local installation or runtime verification was performed. No validation of installation write, build time, language override, impact radius, MCP, graph freshness, token benchmark, visualization, CI, or cloud embedding data path; official benchmarks are not native results. | observed_output: The Latest GitHub Release is v2.3.9, released on 2026-09-18; the official documentation requires Python 3.10 +, the default graph database is located in the project.code-review-graph/graph.db, and the CI documentation indicates that the source code is not sent to external services.
Source: github.com/tirth8205/code-review-graph；https://github.com/tirth8205/c…In contrast to brooks-lint, code-review-graph focuses on extracting symbols, calls, and test relationships from source code to provide agents with queryable structural context; it does not rely on a fixed engineering bibliography to give complete comments. brooks-lint focuses on language model review frameworks, risk classification, and diagnostic formats, but does not persist code diagrams. Use code-review-graph when investigating call chains and impact scopes, and use brooks-lint when unifying the team's architecture and technical debt questioning methods; both require testing and manual review.
Source: github.com/tirth8205/code-review-graph；https://github.com/hyhmrright/…tirth8205/code-review-graph is a project for teams and individuals seeking to enhance code review workflows with AI tools. Its focus on token efficiency and support for a variety of AI coding tools makes it a valuable asset for developers working on large codebases.
The Latest GitHub Release is v2.3.9, released on 2026-09-18; the official documentation requires Python 3.10 +, the default graph database is located in the project.code-review-graph/graph.db, and the CI documentation indicates that the source code is not sent to external services.
The review environment was: Read the tirth8205/code-review-graph official README, v2.3.9 Releases, Usage, FAQ, GitHub Action, VS Code and benchmark instructions; package not installed, local code base not scanned. code-review-graph behavior outside that scope still needs a local check.