New Tool Turns GitHub Repos into Knowledge Graphs
An open-source tool called GitNexus has been released, designed to turn any GitHub repository into an interactive knowledge graph. The tool, which runs entirely in the browser with no server costs, includes an AI agent that allows developers to ask natural language questions about the codebase's structure and relationships.
- The tool was created by developer Abhigyan Patwari and gained over 1,200 stars on GitHub after trending on February 22, 2026. - GitNexus runs a graph database (KuzuDB) and a multi-language parser (Tree-sitter) compiled to WebAssembly, allowing the entire knowledge graph to be built and queried directly in the browser without server-side processing. - Its built-in AI agent uses Retrieval-Augmented Generation (RAG) and features a hybrid search combining keyword (BM25), semantic, and Reciprocal Rank Fusion (RRF) methods for more accurate code retrieval. - The system exposes seven specialized tools to AI coding agents like Cursor and Claude Code, enabling them to perform complex tasks such as impact analysis, coordinated renaming, and automated wiki generation based on the codebase's architecture. - The core problem GitNexus addresses is that many AI coding assistants lack architectural understanding, treating code as flat text files, which leads to errors like missing dependencies or breaking call chains during code generation. - An alternative open-source tool, Fondamenta ArchCode, tackles the same problem by generating structured Markdown files from static analysis, which can be natively read by any LLM without requiring a graph database or a running server. - It supports a wide range of languages including TypeScript, JavaScript, Python, Java, Go, Rust, C, C++, and C#. - This tool is part of a broader movement towards providing deeper context to AI agents, including the Model Context Protocol (MCP), which is also being adopted by tools like Google Chrome DevTools to give agents access to inspection and debugging capabilities.
