# Vectorize releases Hindsight 1.0 memory for agents and launches Memlake storage engine

published: 2026-10-14T07:06:14.772Z
topic: memory
source: https://www.producthunt.com/products/vectorize
url: https://aiagentnews.io/stories/vectorize-releases-hindsight-1-0-memory-for-agents-and-launches-memlake-storage

Vectorize has released Hindsight 1.0, an open-source memory system that lets AI agents learn from past experience. It is also launching Hindsight Memlake, a Rust storage engine for Enterprise customers that keeps agent memory in object storage.

Hindsight is open-source memory for AI agents, and Vectorize says it ranks first on every dataset of the Agent Memory Benchmark. The company calls this version 1.0 and says its API is now a commitment it will keep. Teams can self-host Hindsight for free or start on Hindsight Cloud with free credits.

Memlake is a storage engine written in Rust that keeps agent memory in object storage instead of a database. Vectorize says it handles 100M memories at under 100 ms p99 without a database, consensus cluster or leader. Memlake comes with Hindsight Enterprise, and customers can run it in their own cloud.

In a forum post, Vectorize said recall on a 200k-memory bank took 5.3 seconds on Postgres and 113 ms on Memlake. The company credits turbopuffer's object-storage-first design as an inspiration for the project.

## Why it matters if you build agents

Developers whose agents accumulate large long-term memories can try Hindsight locally on Postgres under an MIT license before choosing a hosted option. The latency figures are Vectorize's own and come from a single test bank, so measure recall on your own data before switching. Memlake is offered to Enterprise customers, so smaller teams should plan around the open-source or cloud tiers.