Graphify converts repositories, documents and images into a queryable knowledge graph that AI coding assistants can explore through Model Context Protocol. The open-source project aims to reduce repeated file reading and help agents follow dependencies that span languages, documentation and configuration.
Its pipeline scans directories, extracts syntax-tree structure with tree-sitter, adds semantic cues from documents and groups related nodes through community-detection algorithms. Recent updates preserve Terraform block attributes, resolve methods across split Rust implementations, track external Kotlin receivers and route scope-qualified C++ calls. Markdown references and scripts embedded in PHP can also be linked into the graph.
The project launched in April and continues to ship frequent updates. Its maintainers report gains on memory and repository question-answering tests, including an increase in key-fact coverage on a small six-question ERPNext evaluation. Those first-party results are too narrow to establish a broad advantage over ordinary search or repository indexing.
Community reports describe strong value for onboarding and architectural exploration, but more mixed results in everyday work. For medium-sized repositories, familiar text search can still be faster than building and querying a graph. Graphify is therefore most relevant when cross-file relationships and mixed documentation create more difficulty than the setup cost it adds.