About

Building the memory layer for AI.

We build the infrastructure that lets AI agents remember, learn, and improve over time.

Mission

We believe AI agents should remember everything they learn. REM Labs builds infrastructure-grade memory primitives — persistent, queryable, and self-improving — so developers don't have to.

What We've Built
  • #1 on LongMemEval (ICLR 2025) with 97.2% accuracy — the highest score on the leading benchmark for long-term memory in language models.
  • 33 framework integrations — LangChain, CrewAI, Vercel AI SDK, AutoGen, and more. Drop-in memory for the tools developers already use.
  • Dream Engine — a 9-stage memory synthesis pipeline inspired by REM sleep neuroscience. Memories consolidate, connect, and strengthen autonomously.
  • Self-hostable via Docker — full control over your data. No vendor lock-in, no usage limits on self-hosted instances.
  • Used by developers building AI agents, coding assistants, research tools, and personal knowledge systems.
Open Source & Standards

We publish the Open Agent Memory Standard (OAMS) and provide SDKs for Node.js, Python, and CLI — so any agent framework can integrate persistent memory without building from scratch.

Contact

For general inquiries, partnerships, or anything else.

hello@remlabs.ai