Weaviate Makes Engram Generally Available, Giving AI Agents Production-Grade Memory

Weaviate, a prominent company known for its open-source AI-native vector database, has announced the general availability of Engram, a new managed memory and context service specifically designed for agentic applications. This launch addresses a critical challenge faced by organizations deploying AI agents in production: the lack of a reliable and scalable memory layer. Traditionally, AI agents struggle with remembering user interactions across sessions, learning effectively from feedback, and maintaining shared context when multiple agents collaborate on a task. Developers often resort to cumbersome workarounds, such as replaying entire conversation histories or manually curating files, which are inefficient and do not scale with the complexity of real-world AI applications. Engram aims to overcome these limitations by treating memory as a fundamental piece of infrastructure, akin to storage and retrieval systems. The service transforms raw agent interactions into structured, durable, and permission-scoped memories, which are then served through Weaviate's powerful vector database. This approach ensures that AI agents have access to a consistent and evolving knowledge base. A key feature of Engram is its use of asynchronous pipelines. These pipelines work in the background to extract relevant facts from agent events, reconcile them against existing knowledge, handle deduplication, and update memories to reflect new information. This "fire-and-forget" mechanism allows applications to hand off raw events and continue processing, while memory is built and maintained durably without introducing latency. Memories stored within Engram are served back through Weaviate's hybrid search capabilities, enabling teams to leverage their trusted retrieval stack for memory management. The service also emphasizes scoping, ensuring that memories are isolated per project, user, and property, thereby delivering the right context to the appropriate caller. Engram is now available in Weaviate Cloud, offering a free tier that includes 1,000 pipeline runs per month, with paid plans starting at $45 per month. Weaviate CEO and co-founder, Bob van Luijt, highlighted that Engram provides a production-grade memory layer, eliminating the need for teams to build and operate one themselves. This simplifies the development of sophisticated AI agents capable of personalization, continual learning, and managing multi-agent shared states through ready-made templates and customizable options.