Hypersave v1.0 GA
Memory that thinks. The cognitive memory layer for AI agents — now generally available.
Today we are shipping Hypersave v1.0 as a generally available product. After nine months in private beta with 40+ teams building agents on top of GPT, Claude and our own AVALON line, the cognitive memory layer is open for anyone to use.
Hypersave is the first memory system designed around how memory actually works — not as a flat vector store, but as a layered architecture of episodic, semantic and procedural recall, with a forgetting curve that compresses old context the way a brain does. Agents built on Hypersave keep what matters and let go of what doesn't.
The numbers from beta are why we're shipping now. Across the cohort, agents using Hypersave reduced average context-window spend by 71% on long-running tasks, while improving recall accuracy on week-old conversations by 38 points over a flat RAG baseline. One design partner — a customer-success agent serving a Fortune 500 SaaS — went from a 4,200-token average prompt to 1,180 tokens, with measurably better personalisation.
v1.0 ships with a managed cloud at platform.hypersave.io, an open-source self-hosted runtime under Apache 2.0, native SDKs for Python and TypeScript, and first-party adapters for the major agent frameworks. Pricing starts free for teams under 100K writes per month. The pricing page has the rest.
Memory is the bottleneck for agents. Hypersave is our answer. Build with it.
Slow updates from the lab.
One email per release. No marketing, no tracking, unsubscribe anytime.