GitNexus is a zero-server code intelligence engine that transforms any codebase into an interactive knowledge graph directly in your browser. Built on KuzuDB compiled to WebAssembly and Tree-sitter WASM for multi-language parsing, it processes GitHub repositories or ZIP files entirely client-side with no data leaving your machine. The engine maps every dependency, call chain, cluster, and execution flow into a navigable graph structure, then exposes this through a built-in Graph RAG agent that can answer complex questions about code relationships, architecture patterns, and cross-module dependencies across 14 programming languages including TypeScript, Python, Go, Rust, and even legacy COBOL.
The tool offers deep integration with AI coding assistants, particularly Claude Code through MCP tools, agent skills, and lifecycle hooks. PreToolUse hooks automatically enrich agent searches with graph context, while PostToolUse hooks reindex after commits to keep the knowledge graph current. Features include community detection for identifying tightly coupled modules, process detection for tracing execution flows, confidence scoring for relationship strength, and multi-repo MCP support for indexing multiple codebases simultaneously. A local backend mode via gitnexus serve provides persistent indexing with full AI chat support through the web UI.
GitNexus has attracted over 24,000 GitHub stars since its release, reflecting strong demand for privacy-preserving code understanding tools. The project ships under a permissive license with zero configuration required for basic use — point it at a repo and the graph builds automatically. Recent updates added mainframe COBOL indexing for legacy system modernization, making it one of the few tools that bridges modern AI code intelligence with enterprise legacy codebases. The browser-first architecture means teams can explore code graphs without installing dependencies, making it particularly useful for onboarding new developers and conducting architecture reviews.