ReMe
ReMe: Memory Management Kit for Agents - Remember Me, Refine Me.
ReMe is a local-first memory management framework that gives AI agents persistent, file-based knowledge storage. It transforms conversations and external resources into structured Markdown memory organized in daily notes and long-term digest layers, supporting self-evolution through automatic consolidation workflows. Agents access memory through CLI commands, HTTP API, MCP server, or native integrations for DeepSeek Harness, OpenClaw, QwenPaw, Claude Code, and other platforms. The system uses BM25 search, optional embeddings, and wikilink graphs to retrieve relevant context without loading entire knowledge bases into prompts.
Key Features
Use Cases
- 01Maintaining long-term conversational memory across multiple chat sessions with personal assistants
- 02Building persistent coding knowledge bases that capture project conventions and technical decisions
- 03Creating self-evolving personal knowledge repositories from conversations and imported documents
- 04Enabling multi-agent workflows where coding agents and assistants share the same memory workspace
- 05Tracking financial research with automatic consolidation of news, analysis, and historical context
- 06Implementing experience-driven tool-use improvement by recording execution outcomes and procedures
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
ReMe — FAQ
What is ReMe and what does it do?+
ReMe is a memory management framework for AI agents that stores conversations and resources as Markdown files in a local workspace. It automatically consolidates daily interactions into long-term knowledge using Auto Memory, Auto Dream, and search workflows, allowing agents to recall and evolve information over time.
How do I install ReMe?+
Install ReMe with pip using 'pip install "reme-ai[core]"' for Python 3.11+, then start the service with 'reme start'. The core installation includes the Studio web interface; embedding support requires additional configuration of LLM and embedding API keys in a .env file.
Which AI agents and clients work with ReMe?+
ReMe integrates with DeepSeek Harness and OpenClaw via the @agentscope-ai/reme npm package, QwenPaw through its Python API, Claude Code via MCP plugin, Hermes through HTTP provider, and CLI agents like Codex using the SKILL.md memory skill. Any agent can access it through the CLI, HTTP API, or MCP server.
Do I need API keys or LLM access to use ReMe?+
Basic file operations, BM25 search, wikilink traversal, and reading proactive topics work without LLM credentials. Auto Memory, Auto Resource, Auto Dream workflows require LLM API keys; embedding-based semantic search requires separate embedding API configuration, but is disabled by default.
Is ReMe free and open source?+
Yes, ReMe is open source under the Apache License 2.0. The framework itself is free; users only pay for their own LLM and embedding API usage if they enable those features.
Where is memory stored and can I edit it manually?+
Memory is stored as ordinary Markdown files in a local .reme/ workspace directory (or custom location). Users can directly read, edit, move, sync, and backup these files with any text editor or file tool; metadata indexes are automatically rebuilt.
How do I install ReMe?+
Open the source repository on GitHub and follow its README. ReMe is a skill — MCP Agents Market links you directly to the official repo.
Is ReMe free?+
ReMe is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.