Safi Shamsi · 2026-04-03 · notable
Graphify — Turn Any Folder Into a Queryable Knowledge Graph
AI coding assistant skill that converts code, docs, papers, images, and videos into queryable knowledge graphs using tree-sitter AST and Leiden clustering. 71.5x fewer tokens per query than reading raw files. 23.2k stars.
Type /graphify in your AI coding assistant and get a queryable knowledge graph of your entire codebase.
Key specs
| License | MIT |
|---|---|
| GitHub stars | 23.2k |
| Languages supported | 22 |
| Token reduction | 71.5x |
What is it?
Graphify is an open-source skill that plugs into AI coding assistants (Claude Code, Codex, Cursor, Gemini CLI, and others). It reads a folder of code, documentation, papers, images, or media files and builds an interactive knowledge graph you can query, rather than having the AI re-read raw files on every question.
How does it work?
A three-pass pipeline: first, tree-sitter extracts AST structure from 22 programming languages without any LLM calls. Second, local Whisper transcription handles audio and video. Third, parallel LLM analysis builds semantic relationships. The result is a NetworkX graph clustered with the Leiden algorithm, output as interactive HTML, JSON, and markdown. Queries hit the graph's subgraph, not the full file set, cutting token usage by 71.5x.
Why does it matter?
AI coding assistants burn tokens re-reading large codebases on every prompt. Graphify collapses that cost by precomputing a structured graph, making deep codebase questions both faster and cheaper. The 23k stars in one week signal real developer demand for this workflow.
Who is it for?
Developers using AI coding assistants on large or unfamiliar codebases.
Try it
pip install graphifyy && graphify install