marm-mcp

# MARM Systems FAQ --- ## 🎯 Base Questions - General MARM ### Q: What is MARM Systems? MARM Systems provides **Universal Memory Intelligence** for AI agents through three main offerings: | Product | Description | Best For | |---------|-------------|----------| | **MCP Server** | Production-ready universal memory server with 19 tools | Claude, Gemini, Qwen, and any MCP-compatible AI | | **Original Protocol** | Copy/paste instructions for manual memory management | Any AI platform (ChatGPT, Claude, local models) | | **Live Chatbot Demo** | Interactive testing environment | Quick testing and feature exploration | ### Q: How is MARM different from built-in AI memory? | Feature | Built-in AI Memory | MARM Systems | |---------|-------------------|--------------| | **Control** | Limited, opaque, no user control | Full user control over what gets remembered | | **Portability** | Platform-locked (ChatGPT only works in ChatGPT) | Cross-platform (memory works everywhere) | | **Validation** | No accuracy guarantees | Built-in validation and reasoning transparency | | **Search** | Basic recency-based | Semantic similarity search by meaning | | **Sharing** | Can't export or transfer | Memory database shared across all AI agents | ### Q: Who is MARM for? **Perfect for:** - **Developers** - Long coding projects requiring context continuity - **Researchers** - Complex analysis with memory accuracy needs - **Enterprise teams** - Shared AI memory across different platforms - **Power users** - Anyone doing serious work with multiple AI agents **Not ideal for:** - Quick, one-off questions - Users wanting fully automated solutions ## 🚀 MCP Server Questions ### Q: How do I install the MARM MCP Server? | Method | Commands | Time | Requirements | |--------|----------|------|--------------| | **Docker (Recommended)** | `docker pull lyellr88/marm-mcp-server:latest`<br>`docker run -d --name marm-mcp-server -p 8001:8001 -v ~/.marm:/home/marm/.marm lyellr88/marm-mcp-server:latest`<br>`claude mcp add --transport http marm-memory http://localhost:8001/mcp` | 2 minutes | Docker installed | | **PyPI Install** | `pip install marm-mcp-server==2.2.3`<br>`marm-mcp-server` | 1 minute | Python 3.10+ | ### Q: What MCP tools does MARM provide? **19 Complete MCP Tools organized by category:** | Category | Tools | Description | |----------|-------|-------------| | **Memory Intelligence** | `marm_smart_recall`, `marm_contextual_log` | AI-powered semantic search and intelligent storage | | **Session Management** | `marm_start`, `marm_refresh` | Memory activation and session state management | | **Logging System** | `marm_log_session`, `marm_log_entry`, `marm_log_show`, `marm_log_delete` | Structured conversation history | | **Notebook Management** | `marm_notebook_add`, `marm_notebook_use`, `marm_notebook_show`, etc. | Reusable instructions and knowledge storage | | **Workflow Tools** | `marm_summary`, `marm_context_bridge` | Context summaries and workflow transitions | | **System Utilities** | `marm_current_context`, `marm_system_info`, `marm_reload_docs` | System status and information | ### Q: Which AI platforms work with the MCP server? **Currently Supported:** - ✅ **Claude Code** - Full integration with CLI command - ✅ **Qwen CLI** - Complete MCP tool access - ✅ **Gemini CLI** - All 19 tools available - ✅ **Any MCP-compatible client** - Universal protocol support **Coming Soon:** - ChatGPT (when OpenAI adds MCP support) - Additional enterprise AI platforms ### Q: How does semantic search work? **Traditional keyword search:** "authentication error" only finds exact matches **MARM semantic search:** "authentication error" finds related memories about "login problems", "user verification issues", "access denied", etc. **Technical details:** - Uses AI embeddings (`all-MiniLM-L6-v2` model) - Vector similarity search finds content by meaning - Global search across all sessions with `search_all=True` - Intelligent auto-classification (code, project, book, general) ### Q: Can multiple AI agents share the same memory? **Yes! This is MARM's key feature:** - **One database** shared across all connected AI clients - **Cross-platform intelligence** - Claude learns from Gemini's conversations - **Collaborative workflows** - Different AIs contribute to same knowledge base - **Session isolation** available when needed - **User-controlled** sharing and memory management