agentic-graphrag

Agentic GraphRAG is a Python framework that links documents, agent activity, and decisions in one graph, so agents can find evidence, continue work across sessions, and revisit decisions when sources change. ^[1] It turns documents into searchable knowledge using a schema—a defined set of entity and relation types—and gives agents tools to follow relationships, read source passages, and cite what they use. ^[2a][2b] For example, a contract-review agent can record a recommendation with contract and policy passages; a later session retrieves the task and decision, and a policy change flags the decision for review. ^[3]