Memory Engine

Name: Memory Engine
Author: SimoneB79

@SimoneB79

About Memory Engine

Atomic memory model — knowledge stored as atoms (facts, decisions, events, preferences, logs, procedures, notes, session messages)

#memory

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "memory-engine": {
      "url": "http://localhost:8085/sse",
      "transport": "sse"
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is Memory Engine?

Memory Engine is an MCP server that implements an atomic memory model for storing and retrieving knowledge as atoms (facts, decisions, events, preferences, logs, procedures, notes, session messages). It uses multi-factor ranking (FTS relevance × confidence × recency × weight), organic decay, and a learning engine to manage memory over time.

How to use Memory Engine?

—

Key features of Memory Engine

Atomic memory model with typed knowledge atoms
Multi-factor ranking combining FTS relevance, confidence, recency, and weight
Organic decay of unused atoms with review flags for critical ones
Learning engine that detects contradictions, gaps, weak atoms, and merge candidates
Session watcher that ingests OpenClaw session messages as atoms with TTL

Use cases of Memory Engine

Automatically ingesting and organizing conversation logs from OpenClaw sessions
Importing existing markdown notes as a one-way sync into the atomic knowledge graph
Graph traversal and relationship-aware retrieval for knowledge exploration
Automatic deduplication and merging of similar atomic memories
Detecting and flagging knowledge contradictions for human review

FAQ from Memory Engine

What is an atom in Memory Engine?

An atom is a unit of knowledge that can represent facts, decisions, events, preferences, logs, procedures, notes, or session messages. Each atom has properties like weight, recency, confidence, and optional TTL.

How does the ranking of atoms work?

Recall uses a multi-factor formula: FTS relevance × confidence × recency × weight. This is not just BM25 but combines multiple signals to prioritize important or timely knowledge.

Can I import my existing markdown notes?

Yes, Memory Engine supports a one-way sync from existing markdown notes. The import is designed for coexistence with, not replacement of, your original files.

What happens to atoms over time?

Atoms undergo organic decay, losing weight if not accessed. Critical atoms can be flagged for review. Additionally, atoms with TTL expire automatically.