Graphify turns codebases into queryable knowledge graphs for AI assistants

Graphify-Labs released Graphify, a CLI tool that integrates with Claude Code, Cursor, Codex, Gemini CLI and 20+ other AI coding assistants. It parses code locally using tree-sitter to generate deterministic knowledge graphs with labeled edges (EXTRACTED or INFERRED), producing queryable graph.json, browsable graph.html, and a GRAPH_REPORT.md. The tool requires Python 3.10+ and supports ~40 languages for cross-file resolution. A proprietary Graphify Cloud service offers continuous graph updates, monorepo support, and third-party connectors.

How Graphify builds and queries code graphs

• Local-first parsing: Code analysis uses tree-sitter AST with no LLM involvement; only docs, PDFs, images and video use a configured model for semantic processing
• Deterministic output: Every edge is tagged either EXTRACTED (explicit in source) or INFERRED (derived by resolution), with no embeddings or vector store
• Three output files: graph.html (interactive browser visualization), GRAPH_REPORT.md (key concepts and suggested questions), graph.json (full queryable graph with schema_version and graphify_version fields)
• Cross-language support: Resolves calls, imports, inherits, and mixes_in across approximately 40 languages via tree-sitter
• Graph algorithms: Uses Leiden community detection for subsystem identification and calculates "god nodes" for most-connected concepts
• CLI commands: graphify explain <node>, graphify path <A> <B>, graphify query "<question>" for scoped subgraph retrieval

Why it matters

The tool addresses growing developer concern about token costs and data privacy when using AI assistants with large codebases, offering a verifiable alternative to opaque RAG systems. Its open-source core and broad assistant support could reduce lock-in to any single AI coding platform.