Graphify iconGraphify

oss Free Star68k

AI coding assistant skill that transforms code, docs, and media into queryable knowledge graphs with 71x fewer tokens than raw file reading

multimodal rag
64.6K GitHub Stars
71.5x Token Savings
YC S26 Backed

Overview

Graphify is an AI coding assistant skill that transforms unstructured files—code, SQL schemas, R scripts, shell scripts, docs, papers, images, and videos—into interactive, queryable knowledge graphs. It uses Claude vision and tree-sitter AST parsing to extract concepts and relationships, then persists them for querying weeks later without reprocessing. The tool achieves 71.5x fewer tokens per query versus reading raw files, making it ideal for large mixed corpora. Works with Claude Code, Codex, OpenCode, Cursor, and Gemini CLI.

The Verdict

Who Should Use Graphify?

Best For

  • Developers using AI coding assistants on large codebases
  • Teams with mixed code + documentation + diagrams
  • Architecture visualization and understanding
  • Semantic codebase search across projects
  • Context-limited AI tools needing efficient retrieval

Not Ideal For

  • Small projects under 1,000 LOC (overhead not worth it)
  • Single-file scripts (no graph structure to exploit)
  • Teams not using AI coding assistants
  • Real-time streaming analysis needs

What's Great

  • 71.5x token reduction vs raw file reading
  • Multimodal: code, PDFs, screenshots, whiteboard photos
  • Persistent graphs—query weeks later without reprocessing
  • Works with Claude Code, Codex, Cursor, Gemini CLI
  • Tree-sitter AST parsing for accurate code understanding
  • Wikipedia-style article generation (--wiki flag)
  • Edge metadata (EXTRACTED vs INFERRED vs AMBIGUOUS)
  • MIT licensed, active development (YC S26)
GitHub README · PyPI

Watch Out For

  • Initial indexing takes time on large codebases
  • Requires Python 3.10+
  • Graph quality depends on source file quality
  • Many optional dependencies for full feature set
  • Relatively new project (April 2026)
GitHub Issues

Pricing

Open Source
Free
MIT License, full features
PyPI Install
pip install
graphifyy (note: double 'y')
View all features & details

Key Features

  • Knowledge graph from any folder
  • Interactive HTML visualization
  • Persistent JSON graph storage
  • Deep mode for aggressive extraction
  • Watch mode for auto-sync
  • Wiki article generation
  • Incremental SHA256 caching
  • Edge confidence metadata

Supported Inputs

  • Python, JavaScript, Go, Rust, Java, C++
  • SQL schemas, R scripts, Shell scripts
  • Markdown, text documentation
  • PDFs, images (via Claude vision)
  • Screenshots, whiteboard photos
  • Videos (with video extra)

AI Integrations

  • Claude Code
  • Codex (OpenAI)
  • OpenCode
  • Cursor
  • Gemini CLI
  • Ollama (local)

Output Formats

  • graph.html — Interactive visualization
  • graph.json — Persistent data
  • GRAPH_REPORT.md — Highlights & questions
  • cache/ — Incremental processing

Real-World Usage

Community Stats

  • 6,500+ forks
  • 329 open issues (active development)
  • Released April 2026
GitHub, June 2026

Technical Details

  • Python 3.10+ required
  • MIT License
  • v0.8.36 current (June 2026)
  • 20+ optional dependency extras
PyPI

How It Compares

Feature Graphify Neo4j Memgraph Vectara
Primary Use AI coding skills General graph DB Real-time graph Enterprise RAG
Token Efficiency 71.5x savings N/A N/A Good
Multimodal Input Code + images + video No No Documents
AI Assistant Integration Claude, Codex, Cursor Manual Manual API
Persistent Storage JSON Native Native Cloud
Self-hosted Yes (local) Yes Yes No
Price Free (MIT) Freemium Freemium Paid

User Reviews

Loading reviews...