#ai-memory

Penfield gives AI agents long-term memory that compounds over time. Instead of starting every session from zero, your agent remembers conversations, learns preferences, connects ideas, and picks up exactly where it left off. Core capabilities: - 16 tools covering memory CRUD, knowledge graph, context management, artifacts, and reflection - Hybrid search with tunable weights — BM25 (keyword), vector (semantic), and graph traversal - Knowledge graph with 24 relationship types across 8 categories (evolution, evidence, hierarchy, causation, implementation, conversation, sequence, dependencies) - Context checkpoints for session handoff between agents or channels - Artifact storage for files, diagrams, and reference docs - 11 memory types intelligent filtering - Personality system with configurable personas via the portal Also available as a Claude Connector, Claude Code integration, native OpenClaw plugin (4-5x faster), and direct API.

Project description 🧠 RABEL MCP Server Recidive Active Brain Environment Layer Local-first AI memory with semantic search, graph relations, and soft pipelines. Mem0 inspired, HumoticaOS evolved. By Jasper & Root AI from HumoticaOS 💙 🚀 Quick Start # Install pip install mcp-server-rabel # For full features (vector search) pip install mcp-server-rabel[full] # Add to Claude CLI claude mcp add rabel -- python -m mcp_server_rabel # Verify claude mcp list # rabel: ✓ Connected 🤔 What is RABEL? RABEL gives AI assistants persistent memory that works 100% locally. Before RABEL: AI: "Who is Storm?" → "I don't know, you haven't told me" After RABEL: You: "Remember: Storm is Jasper's 7-year-old son" AI: *saves to RABEL* Later... You: "Who is Storm?" AI: *searches RABEL* → "Storm is Jasper's 7-year-old son!" No cloud. No API keys. No data leaving your machine. 🛠️ Available Tools Tool Description rabel_hello Test if RABEL is working rabel_add_memory Add a memory (fact, experience, knowledge) rabel_search Semantic search through memories rabel_add_relation Add graph relation (A --rel--> B) rabel_get_relations Query the knowledge graph rabel_get_guidance Get soft pipeline hints (EN/NL) rabel_next_step What should I do next? rabel_stats Memory statistics 📖 Examples Adding Memories # Remember facts rabel_add_memory(content="Jasper is the founder of HumoticaOS", scope="user") rabel_add_memory(content="TIBET handles trust and provenance", scope="team") rabel_add_memory(content="Always validate input before processing", scope="agent") Searching Memories # Semantic search - ask questions naturally rabel_search(query="Who founded HumoticaOS?") # → Returns: "Jasper is the founder of HumoticaOS" rabel_search(query="What handles trust?") # → Returns: "TIBET handles trust and provenance" Knowledge Graph # Add relations rabel_add_relation(subject="Jasper", predicate="father_of", object="Storm") rabel_add_relation(subject="TIBET", predicate="part_of", object="HumoticaOS") rabel_add_relation(subject="RABEL", predicate="part_of", object="HumoticaOS") # Query relations rabel_get_relations(subject="Jasper") # → Jasper --father_of--> Storm rabel_get_relations(predicate="part_of") # → TIBET --part_of--> HumoticaOS # → RABEL --part_of--> HumoticaOS Soft Pipelines (Bilingual!) # Get guidance in English rabel_get_guidance(intent="solve_puzzle", lang="en") # → "Puzzle: Read → Analyze → Attempt → Verify → Document" # Get guidance in Dutch rabel_get_guidance(intent="solve_puzzle", lang="nl") # → "Puzzel: Lezen → Analyseren → Proberen → Verifiëren → Documenteren" # What's next? rabel_next_step(intent="solve_puzzle", completed=["read", "analyze"]) # → Suggested next step: "attempt" 🏗️ Architecture ┌─────────────────────────────────────────────────────────────┐ │ RABEL │ │ Recidive Active Brain Environment Layer │ ├─────────────────────────────────────────────────────────────┤ │ │ │ Memory Layer → Semantic facts with embeddings │ │ Graph Layer → Relations between entities │ │ Soft Pipelines → Guidance without enforcement (EN/NL) │ │ │ │ Storage: SQLite + sqlite-vec (optional) │ │ Embeddings: Ollama nomic-embed-text (optional) │ │ │ │ 100% LOCAL - Zero cloud dependencies │ │ │ └─────────────────────────────────────────────────────────────┘ Graceful Degradation RABEL works with minimal dependencies: Feature Without extras With [full] Text memories ✅ ✅ Text search ✅ (LIKE query) ✅ (semantic) Graph relations ✅ ✅ Soft pipelines ✅ ✅ Vector search ❌ ✅ Embeddings ❌ ✅ (Ollama) 🌍 Philosophy "LOKAAL EERST - het systeem MOET werken zonder internet" (LOCAL FIRST - the system MUST work without internet) RABEL is built on the belief that: Your data stays yours - No cloud, no tracking, no API keys Soft guidance beats hard rules - Pipelines suggest, not enforce Bilingual by default - Dutch & English, more coming Graceful degradation - Works with minimal deps, better with more 🙏 Credits Inspired by: Mem0 - Thank you for the architecture insights! We took their ideas and made them: 100% local-first Bilingual (EN/NL) With soft pipelines With graph relations 🏢 Part of HumoticaOS RABEL is part of a larger ecosystem: Package Purpose Status mcp-server-tibet Trust & Provenance ✅ Available mcp-server-rabel Memory & Knowledge ✅ Available mcp-server-betti Complexity Management 🔜 Coming 📞 Contact HumoticaOS Website: humotica.com GitHub: github.com/jaspertvdm 📜 License MIT License - One love, one fAmIly 💙 Built with love in Den Dolder, Netherlands By Jasper & Root AI - December 2025

xocortex — an MCP server that gives your AI persistent memory across ALL your projects. • Semantic search recalls past solutions by meaning • Knowledge graph connects related insights • 100% local, no API keys Went deep into cognitive science for this: 🔥 Frustration Indexing (Somatic Marker Hypothesis) → Painful debugging sessions get PRIORITY recall. Those 3-hour bugs? Never forgotten. 😴 Sleep/Dream Mechanism → Background consolidation links & deduplicates memories. Like how human sleep organizes memory. Tech stack: • MCP (Cursor/Claude native) • KùzuDB (graph + vector in one embedded DB) • fastembed (local embeddings)

**Context Keeper** 基于LLM驱动的智能上下文记忆管理系统，专为AI Agent提供企业级记忆能力。 **核心创新**：独创"宽召回+精排序"架构，通过两阶段LLM协作实现多维度检索融合（向量+时间线+知识图谱），准确率从传统RAG的45%提升至**75%+**，召回率达**80%+**。 **技术亮点**：完整实现MCP协议（HTTP/WebSocket/SSE三种传输），四维统一上下文模型（Project/Topic/Conversation/Code），支持短期/长期记忆智能转换。 **性能表现**：10000+ QPS并发能力，P95延迟<500ms，用户/工作空间完全隔离。 **开源生态**：MIT许可，Go语言实现，深度集成Cursor/VSCode IDE。让AI助手真正拥有持续记忆能力，告别重复解释，开启智能协作新时代。

A MCP (Model Context Protocol) server providing long-term memory for LLMs

One simply siphons the excess thoughts from one's mind, pours them into the basin, and examines them at one's leisure. It becomes easier to spot patterns and links, you understand, when they are in this form.

Mirror of

MCP server enabling persistent memory for Claude through a local knowledge graph - fork focused on local development