graphify
Turns your whole project into a queryable knowledge graph — ask the graph instead of grepping code
Install
npx skills add https://github.com/graphify-labs/graphify --skill graphifyOverview
From Graphify Labs (YC S26) (★122K, Apache-2.0): type /graphify in your agent and it parses any folder — code, docs, PDFs, screenshots, even video — into a persistent knowledge graph. Code goes through tree-sitter AST: fully local, no LLM, deterministic and reproducible; no vector store, a real traversable graph. Every edge is tagged EXTRACTED (explicit in source) or INFERRED (resolved by graphify), so you always know what was read versus guessed. You get three artifacts: an interactive graph.html, GraphRAG-ready graph.json, and a plain-language GRAPH_REPORT.md — then answer architecture questions with graphify query / path / explain. On the official benchmark, a roughly 1M-LOC ERPNext repo goes from 70.8% key-fact coverage (a grep-and-read baseline) to 82.0% with this single tool added — and building the graph itself costs zero LLM tokens.
Key features
- Local code parsing: tree-sitter AST + call graph, no LLM, no vector store, code never leaves your machine
- Multimodal ingestion: docs/PDFs/images/video get a semantic pass; cross-source edges and community detection
- Three query modes — query / path / explain; --watch or a post-commit hook keeps the graph current automatically
- Full export ecosystem: graph.html, Obsidian vault, --wiki, SVG/GraphML, Neo4j, MCP server
Use cases
- Onboarding onto an unfamiliar repo: build the graph first, then ask about architecture
- Letting agents query the graph instead of reading files — big token savings on large projects
- Dropping papers, screenshots and notes into one folder to grow a queryable personal knowledge base
Anyone who has AI read code daily; complements archify — graphify finds the structure, archify draws the picture. Note the PyPI package is graphifyy (double-y); the skill auto-installs the CLI on first run.
