Graphify (graphify.net, PyPI graphifyy) builds a persistent, queryable knowledge graph from repository source—C# and solution layout via Tree-sitter, plus optional semantic extraction for docs and diagrams. AI coding assistants (Cursor, Claude Code, Codex, and others) use graphify query, graphify path, and graphify explain to traverse structure instead of re-reading raw files on every question.
ConnectSoft pilots Graphify on large ConnectSoft.* libraries and templates (starting with ConnectSoft.Extensions). Outputs such as graphify-out/graph.json and GRAPH_REPORT.md can be committed so teams share one map; local cache and cost files stay machine-specific per our documented workflow.
For day-to-day .NET 10 work we standardize on graphify update . (AST/code-only, no LLM backend). Full semantic graphify extract is optional when rich doc semantics are worth the API cost. Cursor rules and skills wire assistants to the graph without sending raw source to third-party models for structural questions.
If your platform spans dozens of NuGet packages and Factory templates, Graphify is how we keep AI-assisted refactors accurate, cheaper in tokens, and aligned with real architecture—not ad-hoc search.
