CodeGraph RAG: Turn Your Monorepo into a Queryable Knowledge Graph

Modern monorepos are a double-edged sword: everything lives in one place, but no one can keep the whole picture in their head. CodeGraph RAG solves that by indexing your repository into a hybrid knowledge graph — so you can ask questions about your code in natural language and get answers grounded in your actual architecture.

Why a knowledge graph beats plain vector search

Classic RAG splits your code into chunks and stuffs them into a vector database. It answers "where is X used?" poorly, because it has no idea how files relate. CodeGraph RAG combines two engines:

• Memgraph — a graph database storing functions, classes, imports, and calls as nodes and edges.
• Qdrant — a vector database for semantic similarity over code and docs.

The result is a hybrid graph: vector search finds semantically related code, then the graph structure resolves dependencies, call chains, and cross-language references that pure embeddings miss.

Built for multi-language monorepos

Using tree-sitter parsing, CodeGraph RAG fully supports Python, TypeScript, TSX, JavaScript, Rust, Go, Java, C, C++, C#, PHP, Lua, and Dart — with Ruby and Scala on the way. That means a query like "what calls this Python function from the Go service?" actually works across language boundaries.

What it feels like in practice

Point the CLI at a repo, let it parse and index, then ask natural-language questions: "Which services depend on the auth module?" or "Show me the call path from the API gateway to the database layer." Answers come back with file paths and line references you can act on immediately.

Self-host it in minutes

Because it is fully self-hosted and MIT-licensed, CodeGraph RAG keeps proprietary code private — no third-party API ever sees your repository. Spin it up with the official container image, add your repo, and you have a private code-intelligence layer for your whole team.