code-review-graph

Product Introduction

1. Definition: Code Review Graph is a local-first, persistent semantic graph for code intelligence. It operates as a CLI tool and integrates via the Model Context Protocol (MCP), creating a structured knowledge representation of a codebase's architecture, dependencies, and semantics.
2. Core Value Proposition: It exists to eliminate the massive context overhead and associated latency/cost penalties when using Large Language Model (LLM) coding assistants on large or complex repositories. By providing AI agents with a pre-built, queryable map of the codebase, it dramatically improves the accuracy and speed of code generation, review, and refactoring tasks.

Main Features

1. Local-First Semantic Graph Construction: The tool statically analyzes your source code to build a persistent graph database stored locally. Nodes represent code entities (functions, classes, modules, types), and edges define relationships (calls, imports, inherits from). This process uses abstract syntax tree (AST) parsing and static analysis to extract semantic meaning without executing code.
2. Model Context Protocol (MCP) Integration: Code Review Graph exposes its graph query capabilities through the standardized MCP interface. This allows any MCP-compatible AI assistant (e.g., configured within Claude Desktop, Cursor, or Windsurf) to seamlessly query the graph for context, turning the AI into a codebase-aware expert without manual file fetching.
3. CLI for Graph Management & Querying: A command-line interface provides direct control for developers to build, update, and query the graph. Commands allow for incremental updates on file changes, execution of semantic searches (e.g., "find all functions that call this API"), and dependency visualization, enabling use beyond just AI-assisted workflows.

Problems Solved

1. Pain Point: The "context window bottleneck" in AI-powered development. LLMs have limited context windows; feeding entire large codebases is impossible, and manually selecting relevant files is slow and error-prone, leading to incomplete or incorrect AI suggestions.
2. Target Audience: Senior developers and engineering teams working in large-scale monorepos, complex microservices architectures, or legacy codebases. It's particularly valuable for full-stack developers, tech leads managing refactoring efforts, and open-source contributors onboarding to new projects.
3. Use Cases:
   • AI-Powered Code Review: An AI assistant can query the graph to understand the impact of a pull request, identifying callers of a changed function or dependent services without reading all source files.
   • Large-Repository Refactoring: Safely renaming a widely-used function or API by first using the graph to identify all references and dependencies across the entire codebase.
   • Onboarding & Context Discovery: New team members can use semantic queries via CLI or AI to quickly understand architectural patterns and key dependencies.

Unique Advantages

1. Differentiation: Unlike simple file-based search (grep, ripgrep) or in-memory AST tools (like Tree-sitter plugins), Code Review Graph offers a persistent, queryable semantic layer. Compared to cloud-based code intelligence platforms, its local-first architecture guarantees privacy, works offline, and avoids vendor lock-in.
2. Key Innovation: The fusion of a persistent local graph database with the standardized MCP protocol. This creates a durable, shareable "brain" for your codebase that any compliant AI tool can plug into, moving beyond stateless, session-limited context provisioning.

Frequently Asked Questions (FAQ)

1. How does Code Review Graph improve AI code assistant performance? Code Review Graph reduces AI latency and cost by providing precise, graph-based context retrieval instead of forcing the LLM to process irrelevant files. This leads to more accurate code completions, refactoring suggestions, and review comments by ensuring the AI understands architectural dependencies.
2. Is Code Review Graph compatible with my existing IDE or AI assistant? Code Review Graph is compatible with any development environment that supports the Model Context Protocol (MCP), including Claude Desktop, Cursor IDE, and Windsurf. It functions as a standalone CLI and an MCP server, ensuring broad toolchain integration.
3. What programming languages and codebases does Code Review Graph support? The tool relies on static analysis; its current language support is detailed in the official documentation. It is designed to scale with complex, multi-language repositories and monorepos by building a unified graph across different tech stacks.
4. How does the local-first architecture benefit security and privacy? All code analysis and graph storage occur on your local machine or within your private infrastructure. Your proprietary source code never leaves your environment, addressing critical security and intellectual property concerns for enterprise development teams.
5. What is the difference between Code Review Graph and traditional static analysis tools? While both perform static analysis, traditional tools (e.g., linters, complexity analyzers) output reports or rule violations. Code Review Graph outputs an interactive, queryable knowledge graph designed explicitly for programmatic consumption by AI agents and developers through a standardized API (MCP).