A personalized local memory layer for your AI stack.
Remram manages your context and memories more intelligently than session-based chat history.
At its core is the Dream Routine —
also known as REM.
Each day, Remram:
- Reflects on your inputs
- Reconciles corrections
- Updates structured memory
- Prunes noise
- Refreshes relevant artifacts
This is the write path.
REM consolidates.
When your system needs context, Remram serves it through RAM —
the structured recall layer.
- Hybrid retrieval
- Ranked relevance
- Confidence scoring
- Clean context assembly
This is the read path.
RAM remembers.
Remram is not a chatbot.
It is a local-first memory engine that:
- Maintains persistent identity
- Prevents repeated AI mistakes
- Keeps evolving documents coherent
- Promotes durable facts from conversation
- Prunes stale or low-confidence memory
It gives your AI stack continuity.
Remram is designed as a memory layer for OpenClaw-based systems.
remram-os— orchestration and skill integrationremram-encode— REM (consolidation + mutation)remram-recall— RAM (retrieval + read path)moltbox— local deployment infrastructure
OpenClaw handles orchestration.
Remram handles memory.
AI forgets.
It repeats mistakes.
It drifts.
Remram is built to stop that.
REM refines.
RAM recalls.