Turn your codebase into a knowledge graph your agent can query
Graphify is an open-source skill and CLI that maps your entire project — code, docs, PDFs, images — into a knowledge graph. Instead of grepping through files, your AI coding assistant asks the graph: what calls this, what connects A to B, what does everything flow through.
What is graphify?
Graphify (Graphify-Labs/graphify on GitHub, Apache-2.0) became one of the fastest-rising tools in the agent-skills space in October 2026, passing roughly 128,000 stars. It does one thing: it builds a persistent, queryable map of your project that coding agents traverse instead of re-reading raw files.
One command — /graphify . inside Claude Code, Cursor, Codex, or 20+ other assistants — produces an interactive graph (graph.html), a plain-language report (GRAPH_REPORT.md), and the full persistent graph (graph.json).
How it works
Deterministic code parsing — zero tokens
Your code is parsed locally with tree-sitter AST analysis across about 40 languages. No LLM, nothing leaves your machine. This pass extracts hard facts: what calls what, what imports what, what inherits from what.
Model-assisted pass for docs and media
Docs, ADRs, PDFs, screenshots, and diagrams go through your assistant’s model for concept extraction, then merge into the same graph. This is the only step that spends tokens.
Query the graph, not the files
graphify query asks plain-language questions, graphify path A B traces how two concepts connect, graphify explain unpacks one node. Every edge is tagged EXTRACTED (found in source) or INFERRED (derived) — what’s read versus what’s guessed stays visible.
Should you use it?
It’s for you if…
- ✓Your team’s real context lives in docs, ADRs, and RFCs — not just code
- ✓You’re onboarding onto a large or unfamiliar codebase
- ✓Your agent burns tokens re-reading files to answer structural questions
Skip it if…
- ✕Your repo is small enough that grep is already instant
- ✕Nobody will rebuild the graph after refactors — a stale map misleads
- ✕Install with the official command uv tool install graphifyy (double y) — other graphify* packages on PyPI are not affiliated
Go deeper
Frequently asked questions
Is graphify free?
The open-source engine is Apache-2.0 and free — the code graph costs zero tokens to build. Only the semantic pass over docs and media uses your model’s tokens or API key.
How is this different from vector RAG over my code?
Vector RAG retrieves similar-looking text chunks; it doesn’t know that A calls B. Graphify parses the actual structure deterministically — calls, imports, inheritance — and lets you trace paths through it. No embeddings, no vector database, no server.
Which assistants does it work with?
It installs as a skill for Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot, and 20+ other agentic coding assistants on macOS, Linux, and Windows.
Does the graph replace debugging or code review?
No — a map tells you where things are, not why they’re broken. When the graph surfaces a suspicious call chain, pair it with a systematic debugging workflow; when a change touches a highly-connected node, review it with a high-blast-radius checklist. Both are covered below.
Map first, then method.
Give your agent the workflows senior engineers use once the map points somewhere.
Browse the skills catalog