A live context graph is a shared set of governed real-time data products modeling core business entities, with maintained relationships between them. It sits between agent reads and writes, solving the tradeoffs of read-time transformation (token-heavy), write-time transformation (stale or complex), and in-between approaches. Built on incremental view maintenance, it keeps all data products consistent and up to date within seconds. Agents can navigate entity links via MCP or SQL to assemble context autonomously, enabling broader tasks without custom pipelines per use case. Materialize is presented as the engine for this pattern, ingesting operational data via CDC, Kafka, and webhooks and exposing real-time data products. Production examples include a loyalty platform unifying hundreds of microservices and Day AI, an AI-native CRM that updates a graph of business objects in under a second.
