graphify

Graphify is a developer tool that turns a codebase—along with docs, PDFs, SQL schemas, and other files—into a queryable knowledge graph. It is designed for use inside AI coding assistants: once installed, you type /graphify . in Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot CLI, and many other supported platforms, and the assistant can query the graph instead of grepping through source files. How it works: code is parsed locally with tree-sitter AST, so code mapping is deterministic and does not require an LLM. Docs, PDFs, images, and video can be processed with a semantic pass using the assistant's model or a configured API key, but only if the user enables a backend. The system does not use embeddings or a vector store; it builds a real graph with nodes and explained edges. Every connection is tagged EXTRACTED when it is explicit in the source or INFERRED when resolved by graphify, with an AMBIGUOUS tag also mentioned. Core capabilities described in the README include: god nodes for the most-connected concepts, community detection using Leiden, cross-file links across roughly 40 languages through tree-sitter, and commands such as graphify query for plain-language questions, graphify path for tracing connections between two things, and graphify explain for examining a single node. It also extracts rationale comments like # NOTE: and # WHY:, plus ADR/RFC citations, as first-class nodes linked to the code. Output consists of three files: graph.html for interactive browsing, GRAPH_REPORT.md with highlights and suggested questions, and graph.json for programmatic querying. Installation uses the PyPI package graphifyy (note the double-y). The recommended setup is uv tool install graphifyy or pipx install graphifyy, followed by graphify install to register the skill with the AI assistant. Platform-specific installation commands are available for many assistants, and per-project installs can be made with --project. Optional extras cover PDF extraction, Office documents, Google Sheets, video/audio transcription, MCP stdio server, Neo4j and FalkorDB push, SVG export, Leiden community detection, multiple LLM backends (OpenAI, Gemini, Anthropic, Bedrock, Azure, Ollama), SQL schema extraction, PostgreSQL introspection, and additional language parsers such as Terraform, OCaml, Common Lisp, Robot Framework, Pascal, and BYOND DreamMaker. The README also includes benchmark comparisons against other memory/RAG systems and states that code parsing runs fully local with zero LLM credits for the graph build. It emphasizes that graphify is not a vector index and that the project is developed by Graphify Labs, with early access to a hosted platform at app.graphify.com.