What it does

Graphify ingests code, SQL schemas, docs, PDFs, images, even videos, and builds a knowledge graph from them. You type /graphify . in your AI assistant and get three artifacts: an interactive HTML graph, a GRAPH_REPORT.md summary, and a graph.json you can query directly. The tool also installs a “skill” into your assistant so it consults the graph before grepping raw files.

The interesting bit

The platform coverage is almost comically broad — Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Aider, Kiro, “Google Antigravity,” and more. The install command adapts to each assistant’s quirks: hooks for Claude, persistent instruction files for Cursor, $graphify instead of /graphify for Codex. It’s less a graph builder than a polyglot integration layer pretending to be one.

Key highlights

• Optional extras for PDFs, Office docs, video transcription, Neo4j export, Leiden community detection, and multiple LLM backends (Ollama, OpenAI, Bedrock, Gemini, Anthropic)
• Project-scoped installs write under the current directory (e.g., .claude/skills/graphify/) so teams can commit the skill config
• graphify export callflow-html generates Mermaid call-flow diagrams for architecture review
• The README warns against pip install on Mac/Windows due to environment resolution bugs; uv tool install or pipx are strongly preferred

Caveats

• The PyPI package is graphifyy (double-y); other graphify* packages are unaffiliated — a naming landmine
• Leiden community detection only works on Python < 3.13
• Some extras (e.g., BYOND DreamMaker .dm support) may need a C compiler and python3-dev if no wheel matches your platform

Verdict

Worth a look if you live in AI-assisted IDEs and your projects have outgrown context-window grepping. Skip it if you already have a docs/search system your team actually maintains, or if “install a skill into my agent” sounds like too much abstraction to debug.