agentscope-ai/ReMe

ReMe: Memory Management Kit for Agents - Remember Me, Refine Me.

What it solves

ReMe provides a local-first, self-evolving personal knowledge base for AI agents. It solves the problem of agents losing context across sessions or lacking a durable, user-controllable way to store and refine long-term knowledge, preferences, and procedural experiences.

How it works

ReMe implements a "Memory as File, File as Memory" philosophy, storing all durable memory as ordinary Markdown files with frontmatter and wikilinks. It uses a capture-index-consolidate-recall loop:

  1. Capture: auto_memory distills facts from conversations, and auto_resource converts external files into daily cards.
  2. Index: A background watcher ingests Markdown files to create rebuildable BM25 and optional embedding indexes, as well as a wikilink graph.
  3. Consolidate: auto_dream periodically extracts reusable units from recent files to create, refine, or correct long-term knowledge nodes in a "digest" directory.
  4. Recall: Agents retrieve precise, line-level passages and their relationships using hybrid search (BM25 and optional vectors) and wikilink expansion.

Who it’s for

It is designed for developers building AI agents (such as personal assistants or coding agents) who want their agents to have a persistent, transparent, and evolving memory that is shared across different agent runtimes.

Highlights

  • Human-Readable Storage: Memory is stored as Markdown, allowing users to inspect, edit, and back up their agent's knowledge using standard tools.
  • Self-Evolving: Automatically turns conversations and resources into long-term knowledge through a background "dreaming" process.
  • Cross-Agent Compatibility: Works with multiple agent frameworks via Python APIs, HTTP, MCP, or specific plugins for DeepSeek Harness, OpenClaw, and Claude Code.
  • Local-First: Keeps the memory workspace on the user's local machine, ensuring data ownership and control.

Related

  • Project
    NevaMind-AI/memU

    memU is a lightweight, agent‑driven personal memory system that records LLM session logs, turns useful interactions into reusable Markdown “skills”, and makes those skills searchable across sessions, agents, and devices. It works via small host‑adapter binaries (e.g., memu‑codex, memu‑claude‑code) that hook into agents’ session logs and instruction files. The core memory logic is ~500 lines of Python, storing embeddings in SQLite, Postgres, or an in‑memory store. Installation is a one‑message setup with an API key from memU Cloud (or self‑hosted). Once installed, agents automatically retrieve relevant skills before answering, and developers can feed custom conversations into memU via `memu memorize`. Licensed under Apache‑2.0.

  • Project
    doobidoo/mcp-memory-service

    A self‑hosted memory backend for LLM agents that provides a REST API, CLI, OAuth, and a web dashboard. It stores facts, decisions, and causal knowledge graphs locally (SQLite + ONNX embeddings) or in hybrid/cloud mode, offering fast (≈5 ms) semantic search, tag‑scoped retrieval, and optional LLM‑generated summaries. Works with LangGraph, CrewAI, AutoGen, Claude Desktop, ChatGPT, and many other tools.

  • Project
    mem0ai/mem0

    Mem0 is an open‑source memory layer for LLM‑based assistants, providing multi‑level, hybrid retrieval (semantic, BM25, entity linking) and temporal reasoning. It offers a Python/JS SDK, CLI, self‑hosted Docker stack, and a managed cloud service. New 2026 algorithm improves token efficiency and latency, with benchmark scores up to 94.4 on LongMemEval. Ideal for building personalized chatbots, support agents, or any AI system that needs long‑term, user‑specific memory.

  • Project
    grpcer/ownmem

    A Git-native memory system for AI coding agents that stores project knowledge as reviewable Markdown in the repository, providing deterministic local recall without external API calls.

  • Project
    a2328275243/mempalace-evolve

    A local-first long-term memory layer for AI agents that stores durable facts and manages context through isolated project boundaries called wings.