Redis is pushing further into the AI infrastructure race with the launch of Iris, a new context and memory platform designed specifically for AI agents, marking a strategic shift beyond its long-standing role as a high-performance caching layer.
The company said Iris is built to address a growing structural mismatch in modern AI systems: while large language model–driven agents generate vastly more data requests than human users, most existing data retrieval architectures were originally designed for far lower, human-scale demand. As enterprises increasingly deploy autonomous agents, Redis argues that this gap risks becoming a critical bottleneck.
Iris is positioned as the connective layer between AI agents and the data they rely on to function. At its core, the platform combines three capabilities that are typically fragmented across multiple tools. Its Context Retriever enables real-time access to both structured and unstructured data, allowing agents to ground outputs in up-to-date information rather than static training data. The Agent Memory component introduces persistent state, supporting both short-term and long-term recall so agents can maintain continuity across sessions, adapt to user preferences, and execute multi-step workflows more effectively. Meanwhile, the Redis Data Integration (RDI) layer continuously synchronizes and updates underlying data sources, ensuring that retrieved context remains fresh.
The launch highlights a broader industry challenge tied to the stateless nature of large language models. Without external systems to provide memory and context, each interaction effectively resets, limiting the sophistication of AI-driven applications. By consolidating retrieval, memory, and integration into a single platform, Redis is aiming to simplify enterprise architectures while improving performance and reliability.
Alongside Iris, Redis also introduced a new Flex deployment option based on SSD storage. The move signals an effort to balance performance with cost efficiency, as purely in-memory systems—while fast—can become prohibitively expensive at scale. By extending storage to SSDs, enterprises may be able to support larger context windows and longer-lived agent memory without significantly increasing infrastructure costs.
The release of Iris underscores Redis’ ambition to evolve into a foundational layer for agent-based computing, as companies look to operationalize AI systems that require not only fast data access but also persistent, context-aware intelligence.