What is GraphRAG?

GraphRAG is a variant of retrieval-augmented generation that grounds answers in a knowledge graph rather than, or alongside, a flat collection of embedded text chunks. Standard RAG retrieves the passages most similar to a query, which works well for direct lookups but struggles with questions that require connecting facts scattered across many sources or summarizing a whole corpus. GraphRAG addresses this by first extracting entities and the relationships between them from your documents, building a graph, and often clustering it into communities with pre-computed summaries. At query time the system can traverse relationships and pull in connected context, so it answers multi-hop and global questions, who relates to what, how things changed over time, what themes span the dataset, that vector similarity alone misses. Microsoft's GraphRAG project brought the term into wide use, though the underlying idea of graph-grounded retrieval is broader. It is heavier to build and maintain than plain RAG because of the extraction and graph-construction step, so teams often use it where relationships and cross-document reasoning are central. Graph databases exposed as MCP servers let an agent query such a graph directly through tool calls.