code-graph-rag

Why we included this project

Working in a large multi-language repository, keeping track of how modules connect is the hard part, and this project targets that problem directly. It parses source code with Tree-sitter and builds a knowledge graph of the relationships it finds, stored in a graph database. Retrieval-augmented generation runs against that graph, so you can ask natural-language questions and get answers tied to real symbols and their call sites instead of guessed-at file chunks. The same graph-informed context also supports editing and optimizing code, not just search, and the tool runs as a CLI daemon or a Model Context Protocol server, so it can plug into the AI assistants your team already uses. For developers and tech leads who need answers that span many files rather than a scattered file-by-file search, this is a genuinely practical fit.