#USA #San Francisco #AI Agents #Redis Iris #Real-time Context

Redis Iris Enhancements for AI Agents

Redis has recently unveiled the next stage of development for Redis Iris, its sophisticated real-time context engine tailored for AI implementations. This latest update introduces six customizable functionalities, each designed to foster effective communication, context, and memory management, essential for AI agents to function competently in real-world applications. The new features include Redis Intent Router, Redis Agent Playbooks, Redis Agent Memory, Redis Context Gateway, Redis Lens, and Redis Agent Mesh.

Addressing the Challenges of AI in Production

While the advancements in AI models and infrastructure have reached a level of maturity suitable for enterprise deployment, many teams are still facing challenges when it comes to the practical application of AI agents. Often, these agents fall short due to a lack of the right contextual understanding at crucial moments, or they struggle with memory management that doesn’t recall information as required. Furthermore, existing solutions may deliver only marginal value, laden with excessive prompts, redundant model calls, and broken integrations, which can outweigh their productivity benefits.

Rowan Trollope, the CEO of Redis, asserts, “Agents require the same foundational elements used in apps and microservices—live app data, runtime states, and effective service communication. They also require innovations like memory management and contextual routing.” The newly updated Redis Iris aims to eliminate the need for multiple infrastructure products by consolidating these functions into a unified platform, simplifying the development of reliable, production-ready AI agents.

An Overview of Six New Capabilities

Each of the newly introduced features provides unique advantages:

• - Intent Router (General Availability): This capability intelligently interprets the intent of a request rather than merely its wording, thereby directing it to the appropriate destination, whether it be a saved FAQ, an agent skill, a human representative, or an LLM.

• - Agent Playbooks (General Availability): It allows teams to establish stable, reusable answers and simultaneously delineate responses that necessitate up-to-date information and the execution of established procedures. Over time, as playbooks are utilized, they improve in detail and accuracy, leading to faster resolutions and reduced token expenditures.

• - Agent Memory (General Availability): This function forms memories based on user interactions and semi-structured data. Through MemoryETL, it organizes data into hierarchies, affording detailed control and enabling the inspection, auditing, and source tracing of information. It is also equipped to delete unnecessary data, such as personally identifiable information.

• - Context Gateway (General Availability): This serves as a unified access point for agents to manage connected context sources. Teams can implement their own MCP servers and develop tools to optimize problem-solving solutions while promoting consistency across user sessions.

• - Lens (Beta): This feature allows users to trace agent behaviors back to specific skills, routes, or FAQs, providing transparency and actionable insights into task outcomes, resource expenditure for token use per task, and delays encountered.

• - Agent Mesh (Beta): This innovative capability enables agents to communicate seamlessly through temporary streams, ensuring that interactions remain fluid even when connections drop. It implements durable SSE Streams for consistent response delivery and supports reliable collaboration between agents.

These new capabilities enhance the memory and transparency of agents, ensuring they respond accurately and keep track of relevant interactions while maximizing efficiency.

Designed for Real-World Applications

Production-level AI agents must navigate unpredictable human inputs, diversify communication channels, and manage handoffs between various agents; each of these factors carries potential for operational failure. Redis Iris is grounded in four foundational principles:

1. Context must be navigable rather than siloed.
2. Memory should be adaptable and selective.
3. Context efficiency must be prioritized to control costs.
4. Context should evolve as agents learn from previous engagements.

Since its introduction in May 2026, Redis Iris has already been effectively utilized in production environments by numerous customers. For example, June Lee, CTO of Paradigm, explains, “Everything you interact with in Paradigm is stored in Redis. It acts as our memory and orchestration layer, enabling our agents to function in real-time and at an unprecedented scale.”

Availability and Additional Resources

Redis Iris is accessible as of today, with the Intent Router, Agent Playbooks, Agent Memory, and Context Gateway readily available; while Lens and Agent Mesh are currently in beta phase. For further information and resources, interested parties can visit redis.io/iris. Redis will also host a live demonstration featuring the engineering team behind these innovations on November 12 (Americas) and November 17 (EMEA). Registration is open at redis.io/resources/videos/redis-iris-full-spectrum.

About Redis

Redis has established itself as the fastest data platform globally. From its inception as an open-source project in 2011 to becoming the leading brand for caching solutions, it has assisted over 10,000 clients in the development, scaling, and deployment of applications that support contemporary digital experiences. With options for both cloud and on-premises databases geared towards caching, GenAI, and beyond, Redis is setting new standards for performance in application execution. With offices located in cities like San Francisco, Austin, London, Tel Aviv, and Singapore, Redis is recognized as the frontrunner in accelerating app development efficiently. More information can be found at redis.io.