From the source
Harvey is rolling out Harvey Graph, a context layer that maps how a firm's documents, matters, people and legal knowledge connect, so its agents start legal tasks already briefed instead of rebuilding context each time.
- The company says flat retrieval finds passages rather than relationships, forcing agents to re-discover the same entities and cross-references task after task and spend time and tokens recovering structure already in the documents.
- Harvey Graph today combines a knowledge graph covering US Case Law and a per-Space context graph of matters, data and documents from Vaults and integrations, and it follows each Space's governance and security settings without user management.
- Harvey says users will not have to know the graph is there or activate it, but will notice the agent asking less up front while returning work verifiable with citations.
- The company says the investment is far from complete and plans precedent flags, cite checking, law evolution tracking, a new legal research surface and firm-defined legal ontologies in coming months.
- Harvey also says graph value compounds with its memory work, which will accrue to Spaces and stay transparent, traceable and under the user's control.
