Graphify hit 88,467 GitHub stars tagging code without an LLM.

Graphify passed 88,467 stars on GitHub, per its repository.

The tool maps codebases into knowledge graphs without ever calling an LLM, parsing roughly 40 languages locally with tree-sitter AST. No code leaves the machine, which answers the question most teams ask before letting an AI tool near a private repo.

The repository lists the project (Graphify-Labs/graphify) as MIT-licensed, now at version 0.9.17 as of July 16.

Every edge in the graph gets tagged EXTRACTED, meaning it's explicit in the source, or INFERRED, meaning the tool derived it instead. That distinction, not the graph itself, is the whole product: a way to see what was actually read versus what was guessed.

It also runs Leiden clustering to surface god nodes and subsystem communities. The output is a markdown wiki plus an interactive HTML view, and you can query the JSON graph in natural language instead of re-reading files. Install is one line: uv tool install graphifyy && graphify install.

Growth has held near 780 stars a day since April 3, a pace that's stayed steady for three and a half months. My bet: the EXTRACTED/INFERRED split gets copied before the star count does, since honest tagging is harder to fake than a good-looking graph.