graphify

graphify, from Graphify Labs, parses a folder of code into a knowledge graph of functions, classes, files, calls and imports, and answers questions like "what is affected if I change this?" from the graph. It runs as a /graphify skill for Claude Code, Cursor, Codex, Gemini CLI and other agents, or as a command-line tool. With an AI model it can also pull concepts from docs, PDFs and images. Version 0.9.94 shipped October 22, 2026.

How we tested. October 22, 2026, on an 8-core Linux box. We installed graphifyy 0.9.94 from PyPI and ran the code-only path (no AI model, no key) on this website's own 75 TypeScript files. With our tsconfig.json in the folder, it built the graph in half a second and found every file calling each of four functions we checked against a text search (4 of 4, 8 of 8, 3 of 3, 8 of 8), with no wrong files. The answer for one function was 708 bytes against 16,220 bytes for the files themselves. The full test is in the story.

Pro: Fast, local and exact on our code once configured. Answers name the calling function and line, in a fraction of the text an agent would otherwise read.

Con: Without tsconfig.json it couldn't resolve our @/ imports and reported "No affected nodes found" for functions four and eight files call, with no warning. Check one answer against a text search before trusting an empty result.