Redis ships Iris, the memory layer for production agents

verified

Tuesday, May 19, 2026

The News

Redis launched its real-time Context Engine, designed to help enterprise AI agents move beyond simply chatting to users and become productive workers. Three core tools sit behind it — Redis Context Retriever, Redis Agent Memory, and Redis Data Integration — with the latter generally available starting Monday (SiliconANGLE). Redis's flagship in-memory data store already exists in 43% of all enterprise AI agent stacks. VentureBeat's Q1 2026 RAG Infrastructure Market Tracker found buyer intent to adopt hybrid retrieval tripling from 10.3% to 33.3% between January and March, with retrieval optimization surpassing evaluation as the top enterprise investment priority for the first time.

The Read

The frame to hold: enterprises spent 2025 measuring whether their RAG worked, and 2026 paying to fix it. Retrieval optimization investment rose from 19% to 28.9% across Q1, overtaking evaluation spending for the first time. The context layer is an active procurement decision, not a roadmap item. Redis's pitch is that agents need a fridge, not a grocery run on every query — and the real test isn't the metaphor, it's whether MCP-native tool generation actually beats hand-rolled text-to-SQL on agent recall rates in production. If it does, the vector-DB-only vendors that raised at 2024 multiples have a problem: they sold a single primitive into a market that now wants a four-primitive stack from one vendor.

Do This Week

At your next agent-architecture review, ask the engineer who owns retrieval to produce three numbers from last month's logs: retrieval latency p95, percentage of agent stalls attributable to missing context, and percentage of hallucinations traceable to retrieval errors. If your team can't produce those three numbers, you're not in a position to evaluate any context-engine pitch that's about to land on your desk — and several will.

For Product lead