graphify

Project Overview

Graphify is a codebase knowledge-graph tool for AI coding assistants: type /graphify in Claude Code, Cursor, Codex, or Gemini CLI and it parses your entire project — code, docs, SQL schemas, configs, even PDFs — into a knowledge graph using deterministic tree-sitter AST parsing, so agents query the graph instead of grepping through files. It runs fully locally with no LLM calls for code and no vector store, every edge is tagged as source-extracted or inferred, and the run produces three files: an interactive graph.html, a GRAPH_REPORT.md summary, and the full graph.json.

Item	Value
Language	Python
Platform	macOS, Linux, Windows (CLI)
Integrations	Claude Code, Cursor, Codex, Gemini CLI, and 20+ assistants
License	Apache-2.0

Turn a codebase into a queryable, traceable, clickable graph.

Key Features

• Deterministic AST parsing: code is mapped via tree-sitter into a stable, reproducible graph with no LLM in the loop and nothing leaving the machine
• Every edge explained: connections are tagged EXTRACTED (explicit in the source) or INFERRED (resolved by graphify), so trust levels are visible at a glance
• Real graph traversal: no embeddings and no vector store — ask questions, trace paths between two concepts, or explain a single concept
• Interactive visualization: graph.html renders detected communities as a force-directed graph with clickable, filterable, searchable nodes
• Full graph data: graph.json preserves the complete structure for querying without re-reading sources, with schema versioning for integrations
• One-command assistant setup: graphify install detects installed AI assistants and registers the /graphify skill
• Multimodal coverage: beyond code it processes docs, PDFs, images, and video (semantic passes use your assistant's model or a configured API key)

Target Audience

• Developers working heavily with AI coding assistants
• Engineers onboarding into unfamiliar large codebases
• Architecture reviewers and code reviewers
• Teams optimizing agent context efficiency

Quick Start

1. Install the CLI (needs Python 3.10+ and uv or pipx): uv tool install graphifyy
2. Register the skill: graphify install detects the assistants on your machine
3. In an assistant session run: /graphify .
4. Explore the output: open graphify-out/graph.html in a browser, read GRAPH_REPORT.md for highlights, query graph.json programmatically

Core Advantages

• Context savings: agents query the graph instead of searching full text, cutting token spend significantly
• Trustworthy answers: deterministic parsing plus edge-level provenance keeps conclusions verifiable
• Local first: all code parsing happens locally; source code never leaves the machine

System Requirements

Item	Requirement
Language	Python 3.10+
OS	macOS, Linux, Windows
Installer	uv or pipx
Assistants	Claude Code, Cursor, Codex, Gemini CLI, and more

Notes

• PyPI name differs: the install package is graphifyy (double y), while the command and skill are named graphify
• Semantic passes need a model: code graphing is fully local and deterministic; semantic analysis of docs, PDFs, and other non-code content uses the assistant's model or a configured API key