Redis Iris

Redis Iris is a context engine that sits between an AI agent and the enterprise data it needs to act, turning fragmented systems of record into a governed, agent-readable retrieval layer with durable memory. It is aimed at teams running production agents that fail on stale state, slow retrieval and fragmented memory rather than on model quality.

Redis Iris is Redis, Inc.'s context and memory platform for AI agents, announced 18 May 2026 by CEO Rowan Trollope. Redis's argument is that production agents fail for runtime reasons — stale state, slow retrieval, fragmented memory, disconnected tools — rather than because of model choice, and that the fix is a dedicated context engine between the agent and its data. Iris packages five tools on top of Redis. Context Retriever lets developers define a semantic model of business data — entities, fields, relationships and access rules — and auto-generates Model Context Protocol tools the agent calls instead of querying databases directly or gambling on text-to-SQL; agents authenticate with scoped keys, discover only permitted tools, and row-level filters are enforced server-side. Agent Memory manages short-term conversational state plus longer-lived durable memory so preferences and history persist across turns, sessions and agents. Redis Data Integration continuously synchronises relational databases, warehouses and document stores into Redis in agent-optimised formats, separating systems of record from the agent-facing retrieval plane. LangCache provides semantic caching that Redis says cuts token costs by up to 90%, and Redis Search is the underlying vector, structured and unstructured query engine. Context Retriever and Agent Memory launched in preview; Redis Data Integration reached general availability on the same date. Redis claims sub-millisecond reads and writes, under 250ms P95 query latency across production workloads, and that Redis already sits in 43% of enterprise AI agent stacks; it passed $300M in annualised recurring revenue. Iris is multi-cloud and BYOC by design, positioned explicitly against making an agent context layer a hyperscaler lock-in decision.

Category

Infrastructure & Cloud

Pricing

Usage-based, Subscription, Freemium, Contact for pricing

Target Market

CTOs, CIOs, Enterprise Developers, Data Engineers, Platform Engineering Teams

Deployment

Cloud-first, Multi-cloud, Hybrid, Self-hosted, API-based

Founded

2011

Headquarters

San Francisco, United States

Customers

10,000+ Redis customers company-wide; Character.ai and Safe in Home are named in Iris materials

• ✓
  Context Retriever (preview)

  Declares a semantic model of entities and access rules, then auto-generates MCP tools agents call instead of writing queries.

• ✓
  Agent Memory (preview)

  Stores short-term interaction history and long-term durable memory so agents keep state across turns and sessions.

• ✓
  Redis Data Integration (GA)

  Continuously syncs relational databases, warehouses and document stores into Redis in formats optimised for agent access.

• ✓
  LangCache semantic caching

  Caches semantically similar prompts and responses, which Redis says saves up to 90 percent on token costs.

• ✓
  Redis Search query engine

  Retrieves vector, structured, unstructured and real-time data as the fast layer underneath the context engine.

• ✓
  Server-side access enforcement

  Agents authenticate with scoped keys, discover only permitted tools, and row-level filters are applied server-side.

• ✓
  Multi-cloud and BYOC deployment

  Runs across clouds and in the customer's own cloud account so context does not become a lock-in decision.

• •
  Customer support agent with real situational context

  A support bot answering 'why is my order late' traverses customer, order, shipment, ticket and policy entities in one governed flow instead of brittle per-system integrations.

• •
  Voice agents inside a latency budget

  Sub-millisecond retrieval keeps context lookups from breaking turn-taking in a real-time voice interaction where any stall is audible.

• •
  Fraud scoring on live operational state

  Data Integration streams transaction and account changes into Redis so a scoring agent decides against current state rather than a stale batch snapshot.

• •
  Cutting inference spend on repeated queries

  LangCache serves semantically equivalent prompts from cache, which Redis states can reduce token costs by as much as 90 percent.

• •
  Persistent memory across a coding agent fleet

  Redis cites Safe in Home using Agent Memory to hold shared context across multiple coding agents working in real time.

Iris has no separate price list — it is consumed through an existing or new Redis Cloud account, so the published Redis Cloud rates are what you pay: Free at 30 MB, Essentials from $0.007 per hour with a $5 monthly minimum, and Pro from $0.014 per hour with a $200 monthly minimum and the first $200 free. Neither Context Retriever, Agent Memory, LangCache nor Redis Data Integration appears as a separately metered line item on the pricing page, and Blocks & Files noted the launch announcement carried no pricing information at all. On-premise, hybrid and multi-cloud deployments are annual-plan and quote-only. Buyers should assume the AI capabilities are bundled into the underlying Redis Cloud consumption and confirm metering with sales before sizing, particularly for the two preview components.

This page was written from 8 sources, 4 on domains other than redis.io.

1. 1.redis.io — context is all you needvendor
2. 2.redis.io — irisvendor
3. 3.redis.io — pricingvendor
4. 4.redis.io — companyvendor
5. 5.siliconangle.com — redis debuts much needed memory layer enterprise ai agents
6. 6.blocksandfiles.com — 5241795
7. 7.techzine.eu — redis launches context engine for memory ai agents
8. 8.en.wikipedia.org — Redis