Redis expands Iris with six capabilities for production AI agents
Redis has expanded Redis Iris with six customizable capabilities designed to give AI agents memory, context, routing and communication tools for production use. The company says the capabilities can be used together or separately, with four generally available and two in beta.
- Redis Iris now offers six customizable agent capabilities.
- Intent Router, Agent Playbooks, Agent Memory and Context Gateway are generally available.
- Lens and Agent Mesh are available in beta.
- Redis says Iris has been used in production by customers since May 2026.
- Live presentations are scheduled for 12 November and 17 November.
Six capabilities in one execution environment
The updated Redis Iris includes Redis Intent Router, Redis Agent Playbooks, Redis Agent Memory, Redis Context Gateway, Redis Lens and Redis Agent Mesh. Redis describes Iris as a real-time context engine for AI agents, built to provide production-ready functions for context, memory and messaging.
Intent Router and Agent Playbooks are generally available. Intent Router interprets the meaning of a request instead of relying only on its wording, then directs it to an appropriate destination. That destination can be a frequently asked question stored in memory, an agent skill, a human contact or an LLM. It also supports routing to agents and models, guardrails and conditional fuzzy “if” instructions.
Agent Playbooks help teams separate stable responses that can be reused from requests requiring current data and the rerunning of a known procedure. According to Redis, playbooks become more complete and precise as they are used, helping reduce resolution times and token consumption.
Memory, context and communication for agents
Agent Memory, also generally available, builds memory from user actions, documents and other semi-structured data through MemoryETL. It organizes memory hierarchically for more precise control and includes inspection, auditing and source-tracing functions. Redis says the capability can also forget information that should not be retained, including personally identifying information.
Context Gateway provides a single access point to connected context sources, alongside governance mechanisms. Teams can use their own MCP servers, cache information about tools that can solve a given problem to reduce token consumption, and produce reusable results for consistency across sessions.
Lens is in beta and makes agent behavior traceable. Each result can be linked to the skill, routing path or frequently asked question that produced it, as well as the memory items used for learning and the originating session. Redis says Lens provides audit evidence and visibility into task success, tokens per task and latency.
Agent Mesh, also in beta, lets agents communicate through ephemeral streams. It manages buffering and reconnection after a connection failure. Durable SSE Streams are designed to maintain response transmission through interruptions, while A2A Comms provides a channel for agents to delegate tasks and hand work to one another.
Redis targets the challenges of production deployment
Redis says many teams still struggle to operate agents consistently at the level required in production. The company points to overloaded prompts, repeated LLM calls and fragile custom integrations as common ways to deliver value inefficiently, using more tokens and time than necessary.
According to Redis CEO Rowan Trollope, agents need live application data, execution state and communication between services, along with capabilities such as memory, context management and routing. “Developers should not need five infrastructure products to run their agents,” Trollope said in the company announcement. Redis says Iris brings these primitives together in one place, reducing the custom integration work commonly needed to connect information retrieval, memory, routing and messaging.
Iris is based on four principles described by Redis: context should be navigable rather than siloed, memory should be adaptive and selective, context should be economical to control costs, and context should grow as agents learn from past interactions. Redis said Iris has been used in production by customers since its launch in May 2026. Paradigm CTO June Lee said the company stores everything visible in Paradigm in Redis, which acts as the memory and orchestration layer for its real-time agents.
Availability and upcoming presentations
Redis Iris is available today. Intent Router, Agent Playbooks, Agent Memory and Context Gateway are generally available, while Lens and Agent Mesh are offered in beta. Redis is directing users to redis.io/iris for more information.
The company will present the updated version with the engineering team behind it in live sessions on 12 November for the Americas and 17 November for Europe, the Middle East and Africa. Registration is open through Redis's event page.
For developers and businesses building AI agents, the announcement positions Iris as a consolidated infrastructure layer for memory, context, routing, observability and agent-to-agent communication. The source supports Redis's claim that this approach can reduce custom integration work and improve visibility into agent behavior, but does not provide independent performance figures.
