Graphify Turns Codebases Into Knowledge Graphs, Cuts Tokens 71x

Graphify is an open-source tool that hit 86k+ GitHub stars in three months, turning any folder of code, docs, or images into a queryable knowledge graph. It uses tree-sitter AST parsing to analyze code locally without an API key, with optional LLM-based semantic extraction for documentation.

In a small test, a 4-file codebase produced 15 nodes, 3 communities, and 17 relationships, correctly clustering functions into their logical architectural groups and flagging the most connected 'god nodes.' On a larger 49-file benchmark corpus, the tool extracted 285 nodes and 340 edges across 53 communities, reporting a 71.5x reduction in tokens needed per query compared to reading raw source files.

Graphify can also merge graphs from multiple microservice repositories to reveal cross-repo dependencies, and runs as an MCP server so AI coding agents can query the graph in real time instead of reading files directly. For engineers and AI tools working with large, distributed codebases, this offers a measurable way to cut context overhead.