Graphify

I’m seeing Graphify-Labs/graphify trending, and it’s a tooling-focused approach to “codebase understanding” for agent workflows.

What shipped: Graphify claims you can ingest a repository and produce a queryable knowledge graph where edges are explained. The notable constraint is that it avoids a vector store and instead relies on local, deterministic parsing (AST-based where applicable). The repo also describes a “/graphify skill” meant to plug into agent environments (Claude Code, Cursor, Codex, and Gemini CLI).

Where it’s useful (concrete situation): when I need an agent to answer repo-specific questions reliably—e.g., “Which SQL tables does this service read, what constraints exist, and where is the migration logic?”—and I don’t want the usual embedding drift or opaque similarity matches. A deterministic graph built from schemas and docs gives the agent a structured surface to traverse, and I can audit how it arrived at an answer.

What to look at first: - The /graphify skill docs to understand the expected interface for Claude Code/Cursor-style tools. - The ingestion/parsing pipeline (especially how it handles SQL schemas and PDFs). - Any examples showing “explained edges” or traceability in the generated graph.

If you’re building solo-dev or small-team tooling around RAG/agents, this is the kind of infrastructure that can make answers more testable, because the graph comes from parseable artifacts rather than probabilistic retrieval.