#USA #San Francisco #AI Agents #Redis Iris #Context Management

Redis Unveils Redis Iris: Next-Gen AI Agent Support

Redis has launched an advanced version of Redis Iris, its real-time context engine designed specifically for AI agents. This new iteration introduces six customizable features that can be deployed together or independently, empowering agents with the memory, context, and messaging functionality essential for reliable real-world operations.

The AI models and their supporting infrastructures have evolved to a level where they can be used effectively in corporations. However, many teams still struggle to achieve consistent success with agents in production. Often, agents lack the precise context at the right time and an adequate memory. Valuable agents tend to operate inefficiently, relying on overloaded prompts, redundant large language model (LLM) calls, and fragile, bespoke integrations, where the effort involved surpasses the benefits.

Rowan Trollope, CEO of Redis, emphasizes, "Agents require the same foundational elements as applications and microservices—current application data, runtime state, and service communication." Additionally, they necessitate new functionalities such as memory, context management, and routing. As a leading real-time database for applications, Redis is already a part of the mission-critical processing path. Developers should not need five infrastructure products to make agents functional. Iris brings these foundational components into one place, enabling developers to create agents that perform seamlessly in production environments.

Six Features, One Runtime Environment

Iris delivers production-ready context, memory, and messaging capabilities from day one, while allowing teams to tailor each function to their agents. The release comprises:

• - Intent Router (Generally Available): This feature recognizes the intent behind a query rather than sticking to its wording and redirects it to the appropriate destination—be it a saved FAQ response, an agent skill, a human, or an LLM. Furthermore, it supports routing to agents or models, guardrails, and "fuzzy" if-statements to allow for decision-making beyond binary outcomes.

• - Agent Playbooks (Generally Available): With this, teams can designate which responses should remain valid and reusable, and which require current data and re-execution of a known procedure. With every execution, playbooks become more comprehensive and precise, thereby reducing resolution time and lowering token consumption.

• - Agent Memory (Generally Available): This feature generates memories from user actions and documents, amongst other semi-structured data, through MemoryETL. These memories are organized hierarchically for finer control and provide insights and audits as well as provenance. It also offers the capability to selectively forget information that should not be retained, such as personal data.

• - Context Gateway (Generally Available): This provides agents with centralized access to linked context sources under clear governance guidelines. Teams can integrate their MCP servers, cache tools that solve particular issues to reduce token consumption, and create reusable outcomes that ensure consistency across sessions.

• - Lens (Beta): This feature makes agent behavior traceable. Each outcome can be traced back to the utilized skill, route, or FAQ, as well as the memory from which the agent learned and the initial session. This results in verifiable evidence and insights into task success, token consumption per task, and latency.

• - Agent Mesh (Beta): This allows agents to communicate over ephemeral streams, automatically buffering data during disconnections and restoring connections. Durable SSE streams maintain the flow of responses even during interruptions, providing agents with a reliable channel for delegating and handing over tasks.

Together, these features ensure agents retain appropriate information, provide correct answers, and remain auditable with every action, enhancing results per token. Built on familiar Redis technology, companies can minimize the custom integration workload typically needed to connect information retrieval, memory, routing, and messaging.

Designed for Real-World Challenges

Production agents must handle unpredictable human inputs across multiple channels and transitions between agents. Each of these elements presents potential points of failure. Iris is based on four principles: context should be navigable rather than confined within silos, memory must be adaptable and selective, context should be utilized efficiently to control costs, and context's value should increase as agents learn from past interactions.

Since the introduction of Redis Iris in May 2026, customers have been deploying the solution in productive settings. June Lee, Chief Technology Officer at Paradigm, states: "Everything you see in Paradigm is stored in Redis. Redis is the memory and orchestration layer that allows our agents to operate in real time—at a scale previously deemed impossible."

Availability and Additional Information

Redis Iris is available starting today, with Intent Router, Agent Playbooks, Agent Memory, and Context Gateway fully operational; Lens and Agent Mesh are in their beta phase. More details can be found at redis.io/iris. Redis will host an initial live demonstration of the new version on November 12 for North and South America and November 17 for EMEA. Registrations can be made at redis.io/resources/videos/redis-iris-full-spectrum.

About Redis

Redis is the fastest data platform in the world. Since its inception as an open-source project in 2011, Redis has evolved into the most recommended brand for caching solutions, assisting over 10,000 customers in developing, scaling, and deploying applications that underpin our digital world. With cloud and on-premises databases for caching, generative AI, and many more applications, Redis helps digital businesses set new benchmarks for application speed. With offices in San Francisco, Austin, London, Tel Aviv, and Singapore, Redis is recognized internationally as a leading provider for rapidly developing applications in a short timeframe. For more information, visit redis.io.