Redis introduces the Context Engine, a new layer that provides AI agents in enterprise environments with reliable memory. The company aims to address the “context problem” with this solution: a lack of memory that causes autonomous systems to hallucinate, stall, or produce incorrect results.
The Context Engine consists of three components. The Redis Context Retriever, currently in preview, allows developers to build a semantic model of business data. This enables agents to map how customers are related to opportunities or support tickets. Instead of relying on unpredictable text-to-SQL queries, the Retriever automatically generates tools based on the open-source Model Context Protocol.
The second component, Redis Agent Memory, offers a two-tier memory system. Short-term interaction history is managed alongside a persistent long-term cache for preferences and previous sessions. Both are available in preview today.
The third, Redis Data Integration, is generally available today. It continuously synchronizes business data from relational databases and data warehouses, ensuring agents always work with up-to-date information.
Redis’s in-memory datastore is already present in 43 percent of all enterprise AI agent stacks. The company also recently surpassed the $300 million mark in annualized recurring revenue.
With the Context Engine, Redis positions itself as a dedicated infrastructure layer between the agents and the business data.
Tip: Redis 8.2 available with significant performance improvements