MariaDB adds agent memory and a semantic cache to its platform
Enterprise Platform 2027 bundles a durable state store for AI agents and a cache that skips repeat model calls, alongside a self-reported throughput jump over the last major version.
MariaDB released Enterprise Platform 2027 on 4 November, bundling its transactional database, real-time analytics, vector search and a new set of AI tools onto one platform that runs on premises, on MariaDB Cloud, or bring-your-own-cloud.
Two features are aimed squarely at teams building agents. Agent Memory gives an agent a durable, in-memory store for its state, saved after every step and shared across agents and sessions, so a long-running agent can resume work instead of starting over. Semantic Cache recognises when a new question means the same thing as one it has already answered and returns the stored answer without calling the model again. MariaDB AI Retrieval, a separate tool for pulling context into those agents, is in public preview on MariaDB Cloud rather than generally available. The company also claims Enterprise Server 12.3 delivers 56% higher peak throughput than version 11.8.
Why it matters: if you’re already running MariaDB and building anything agentic, this removes a reason to bolt on a separate vector store, cache layer or state store. It’s also a sign that “agent memory” is becoming a database feature, not just an application-layer concern.
The caveat: the throughput and performance numbers are MariaDB’s own sponsored testing, not an independent benchmark, and the AI Retrieval piece isn’t GA yet.