FAQ
FreshGeneral
What is Graphify?
Graphify is an open-source skill for AI coding assistants that transforms any folder of code, documentation, papers, images, or videos into an interactive, queryable knowledge graph. It dramatically reduces the tokens your AI assistant needs to understand complex codebases — by 71.5x on mixed corpora compared to reading raw files.
Why is the PyPI package named graphifyy?
The package name graphify on PyPI is taken by an unrelated project. The Graphify team is working to reclaim it. Until then, install with pip install graphifyy (double-y). The CLI command you use after install is still graphify (no double-y).
Is Graphify free?
Yes. Graphify is open-source and free to use. The only costs are from your own LLM API usage (for prose/doc extraction) and your AI coding assistant subscription.
Does Graphify send my code to external servers?
No. Code files are analyzed via tree-sitter AST entirely locally — no code is sent externally during AST extraction. Only prose/documentation files are sent to your platform's LLM API using your own API key. Video and audio transcription runs locally via faster-whisper.
Building Graphs
How long does the first build take?
It depends on codebase size and the proportion of prose/documentation files:
- Code-only (1,000 files): 1-3 minutes (pure tree-sitter, no LLM)
- Code + docs (1,000 files + 100 markdown files): 5-10 minutes
- Heavy documentation (PDFs, large markdown): 10-30+ minutes
Use .graphifyignore to exclude files you don't need in the graph.
How often should I update the graph?
Run /graphify . --update after significant code changes. For active development, use --watch mode. For teams, use Git hooks (graphify hook install) to update automatically on every commit.
Can I build a graph of a public GitHub repository?
Clone the repo locally first, then run /graphify . in the cloned directory. For remote content like papers or videos, use /graphify add <url>.
What are "God Nodes"?
God Nodes are the highest-degree nodes in the graph — the concepts with the most connections. They represent the core abstractions your system is built around. Reviewing them first is the fastest way to build a mental model of an unfamiliar codebase.
Queries
Are queries case-sensitive?
Query strings for /graphify query are not case-sensitive. Node names for /graphify path and /graphify explain are case-sensitive.
How is /graphify query different from asking my AI assistant directly?
/graphify query searches the persistent graph structure — it finds related nodes and edges without reading any source files. Your AI assistant then uses those results to guide which files to actually read. This is faster and uses far fewer tokens than asking the assistant to search files directly.
Integrations
Can I use Graphify with multiple AI assistants at once?
Yes. Each platform has its own install command. You can have Claude Code, Cursor, and Codex all referencing the same graphify-out/graph.json simultaneously.
Does the MCP server work with Claude Desktop?
Yes. Add the server config to claude_desktop_config.json and point it at your graph.json. See the MCP Server SOP.
Teams
Who should run the initial graph build?
Any team member can run it. The convention is to have one person build and commit graphify-out/ to the repository. After that, Git hooks keep it current automatically.
Should graphify-out/ be committed to Git?
Yes, for teams. Committing the graph lets every team member get instant graph context after git pull. Exclude graphify-out/cache/ in .gitignore — each machine rebuilds the cache locally.
What if two developers update the graph simultaneously and create a merge conflict?
Treat graph.json as a derived artifact. Accept either version and run /graphify . --update to merge current state. Do not manually merge the JSON.