Engram, Weaviate's managed memory and context service for AI agents, is now generally available. It addresses three core failure modes in production agents: long-context degradation, messy raw data accumulation, and multi-agent context fragmentation. Engram runs asynchronous background pipelines to extract, deduplicate, and reconcile agent interactions into structured, durable memories, served via Weaviate's hybrid semantic and keyword search. It ships with ready-to-deploy templates for personalization, continual learning, and multi-agent state, while allowing teams to drop into lower-level pipeline primitives. Built-in scoping provides per-project, per-user, and per-property memory isolation. Engram is available now on Weaviate Cloud with a free tier.
