code-review-graph

code-review-graph is a local persistent code graph tool for AI code review scenarios, with its official website at https://code-review-graph.com/。根据官方提供的资料，它的核心目标是通过代码图谱为 It provides more efficient context for AI code review, thereby reducing token consumption.\n\nIn terms of purpose, this tool mainly serves the AI code review process. It builds and persistently stores a local code graph, so that AI does not need to repeatedly read large amounts of raw files when reviewing code, but instead obtains structured information through the graph. The official claim is a median context reduction of about 63x (~63x median context reduction), meaning that compared with directly providing raw code context, the required context size is significantly reduced after using this tool.\n\nIn terms of main capabilities, the official information explicitly mentions two points: first, it provides 30 MCP tools (30 MCP tools), which can be integrated with AI tools or clients via the MCP protocol; second, it provides CLI command-line access. The tool is local-first by default (local-first by default), with the code graph stored locally, making it suitable for scenarios with requirements for code privacy and local operation.\n\nIn terms of usage, the official materials only state that it can be used via MCP and CLI, and emphasize local-first. Details such as specific installation commands, configuration steps, supported programming languages, which AI clients or editors it integrates with, and the graph generation and update mechanisms are not yet provided in the available materials and should be based on the official documentation.\n\nIn terms of applicable scenarios, it is suitable for developers or teams that need AI-assisted code review and want to control context size and token costs, especially workflows that focus on localized code processing and do not want code to be transmitted externally.\n\nIn terms of limitations, the available materials do not provide information such as pricing models, prices, platform support scope, or detailed test conditions for performance benchmarks, and are not yet provided.