Graphify is an advanced AI coding assistant skill that transforms any directory of code, documents, images, and multimedia files into an interactive, queryable knowledge graph. By mapping out a project’s structure, it allows developers and their AI assistants to instantly understand codebases, uncover the “why” behind architectural decisions, and navigate by structural relationship rather than simple keyword matching.
How It Works
The tool operates in three distinct passes to build its graph. First, a deterministic Abstract Syntax Tree (AST) pass extracts structures from code files without utilizing a Large Language Model (LLM). Second, any video or audio files are transcribed entirely locally using faster-whisper. Finally, LLM subagents run in parallel over text documents, media transcripts, and images to extract semantic concepts and design rationale. The results are merged into a NetworkX graph and clustered using Leiden community detection—relying entirely on graph topology and edge density rather than vector embeddings.
Key Features and Integrations
Graphify is heavily optimized for token efficiency, demonstrating up to a 71.5x reduction in token usage per query by allowing the AI to read a compact graph summary instead of raw files. It natively supports 25 programming languages and integrates seamlessly across multiple developer platforms, including Claude Code, Codex, Cursor, and GitHub Copilot. By utilizing “always-on” hooks, assistants are instructed to consult a one-page graph report detailing “God nodes” (high-degree central concepts) and surprising code-to-documentation connections before executing standard file searches. Additionally, every inferred relationship includes a confidence score to distinguish explicit codebase extractions from AI assumptions.
Privacy and Future Ecosystem
The system is designed with strong privacy principles: AST code parsing and media transcriptions run entirely locally, meaning proprietary code and audio never leave the machine. Only non-code text or images are sent to the user’s configured LLM API. Looking ahead, the creator is leveraging this technology to build Penpax, an upcoming enterprise layer that scales Graphify’s capabilities from single code repositories to an always-on, on-device digital twin mapping a user’s entire digital working life.
Mentoring question
How could adopting a deterministic, topology-based knowledge graph change the way your team uses AI assistants to navigate and onboard onto complex legacy codebases?