SAN FRANCISCO, November 6, 2026 / PRNewswire / -- Redis Today, we announce that the new version of our real-time context engine Redis Iris for AI agents has officially been released. Redis Iris now offers six customizable features that can be used in combination or individually, designed to provide agents with the memory, context, and messaging capabilities necessary for reliable operation in a production environment: Intent Router, Agent Playbooks, Agent Memory, Context Gateway, Lens, and Agent Mesh.

The AI model is now available for enterprises to use, but few teams are able to make the agent run stably in a production environment. Agents often lack the appropriate context, are unable to obtain information at the right time, and the information is not properly stored in their "memory." Additionally, the model's efficiency is not high; issues such as prompt overload, repeated calls to large language models (LLM), and inadequate custom integrations result in the token costs and time expenses outweighing the value it creates.

"Developers shouldn't need five different infrastructure products just to get proxies working. Iris integrates these basic capabilities into one place, allowing developers to build proxies that can run in a production environment," said Rowan Trollope, the CEO of Redis. "Proxies require the same foundations as applications and microservices: real-time application data, execution status, and inter-service communication. They also need new capabilities such as memory, context management, and routing." He also noted, " Redis has already taken on a key role as a reference database for application execution, occupying a crucial position in the processing path."

Six functions, one engine

Iris has provided context, memory, and messaging functions that can be used directly in production from day one, while also allowing teams to customize them for their respective agents. This version includes:

• Intent Router ( GA ): Understand the meaning of a request, rather than its form of expression, and route it to the appropriate target, whether it is a frequently asked question answer stored in memory, a skill, manual processing, or LLM. It also supports proxy and model routing, guardrail mechanisms, as well as fuzzy if conditional statements, allowing decisions to be no longer limited to true or false.
• Agent Playbooks ( GA ): Helps the team determine which responses are stable and reusable, and which ones require the latest data to re-execute the established process. With each execution, playbook becomes more complete and accurate, thereby reducing resolution time and lowering token costs.
• Agent Memory ( GA ): Uses MemoryETL to generate memory entries from user behavior, documents, and other semi-structured data, and organizes them into a hierarchical structure for more fine-grained control; it also provides features for checking, auditing, and tracing the source. It can also forget content that should not be retained, such as personal identification information.
• Context Gateway ( GA ): Provides a single entry point for proxies to access connected context sources and is subject to governance mechanisms. Teams can use their own MCP servers to cache which tools can solve specific problems in order to reduce token consumption and generate reusable results, thereby maintaining consistency across sessions.
• Lens ( beta ): Used to track proxy behavior and provide traceability. Each result can be associated with the skill, path, or FAQ response that generated it, as well as the memory entries and source sessions from which the proxy learned, thereby providing auditable evidence, as well as visual information such as task success rate, the number of token used per task, and latency.
• Agent Mesh ( beta ): Allows proxies to communicate via temporary streams, and automatically handles buffering and reconnection in case of connection interruptions. SSE Streams can maintain the response stream during interruptions, while the A2A Comms protocol provides a reliable channel for proxies to delegate tasks and transfer them to other proxies.

Redis indicates that by combining these functions, it helps proxies to remember relevant information, respond appropriately, and maintain transparency, while achieving more results with less token. Since Iris is built on the Redis technology that the team is already familiar with, enterprises can reduce the integration work typically required for connecting information retrieval, memory, routing, and message delivery.

Designed for the complexity of the real world

In a production environment, agents must handle unpredictable user input, multiple channels, and task handover between agents; any of these can become a point of failure. Iris is based on four principles: context should be explorable, not isolated; memory should be adaptive and selective; context should be used efficiently to control costs; and it should be continuously enriched as agents learn from past interactions.

Since its initial release by Redis Iris in May 2026, customers have been using it in production environments. Redis indicates that San Francisco-based startup Paradigm – which Fortune described as a company attempting to replace traditional spreadsheets with AI proxy clusters – has built the entire platform on top of Redis.

"Everything you see in Paradigm is stored in Redis. It serves as the memory and orchestration layer that supports the real-time operation of our agents, and such a scale was impossible to achieve in the past," said June Lee, the Chief Technology Officer of Paradigm.

Availability and Resources

Redis Iris is available from today. Intent Router, Agent Playbooks, Agent Memory, and Context Gateway have entered the official availability phase; Lens and Agent Mesh are still in the beta phase. For more information, please visit redis.io / iris.

Redis will hold webinars regarding Redis Iris on November 12th for the American region, and on November 17th for the EMEA (Europe, Middle East, and Africa) region, with the engineering team that developed the product attending. Registration is now open at redis.io / resources / videos / redis-iris-full-spectrum.

About Redis

Redis claims to be the fastest data platform in the world. Starting from its open-source initiative in 2011, it has become the most frequently cited brand in caching solutions. To date, Redis has helped over 10,000 customers build, scale, and deploy applications that support the global operations. Redis provides cloud and local databases for caching, generative AI, and other purposes, helping digital enterprises set new speed standards for their applications.

Redis is headquartered in San Francisco and has operations in Austin, London, Tel Aviv, and Singapore. It enjoys an international reputation for rapidly developing high-performance applications. For more information, please visit redis.io.