MemPalace/mempalace
MemPalace is an open-source, local-first AI memory system that stores and retrieves conversation history verbatim using semantic search, supporting pluggable backends like ChromaDB and integrating ...
Awesome MCP › Knowledge & Memory
varun29ankuS/shodh-memory
Shodh-Memory is a persistent memory system for AI agents, robots and edge devices, shipped as a single Rust binary with no external database or cloud service. It is reachable over MCP and HTTP for chat agents and over Zenoh and ROS2 for robotics, and is published to the MCP registry, crates.io, npm as @shodh/memory-mcp, PyPI and Docker Hub. Its central design claim is that no LLM sits in the loop: storing and recalling a memory make zero model API calls. Embeddings come from a local MiniLM model (roughly 22MB, INT8) run through ONNX Runtime; named entity recognition uses a local GLiNER span typer with a schema of 141 fine and 18 coarse entity types, downloaded on first run with a rule-based fallback. Typed relations such as LocatedIn, WorksAt and Causes are extracted with directed lexical cues and exemplar-matched semantic typing, building a knowledge graph that supports causal lineage queries by walking causal edges backwards rather than asking a model. Memory dynamics are mathematical: Hebbian strengthening makes frequently used memories easier to find, activation decay fades unused ones, spreading activation surfaces related memories on recall, and long-term potentiation makes repeatedly used connections permanent. Memories move through three tiers, from a 100-item working memory to a 100MB session memory to long-term storage in RocksDB, following Cowan and Wixted memory models. A unified CLI covers init, server (port 3030), TUI dashboard, hook setup, status and doctor. Integration paths include claude mcp add with the npm package plus optional Claude Code hooks for automatic session capture, a Cursor and Claude Desktop JSON config, Docker, and Python and Rust libraries. The README compares it against mem0, Cognee and Zep on LLM calls, latency and offline operation.
https://github.com/varun29ankuS/shodh-memory
Also in Knowledge & Memory
siyuan-note/siyuan
SiYuan is a privacy-first, self-hosted personal knowledge management system featuring block-level referencing, Markdown WYSIWYG editing, and support for the Model Context Protocol (MCP).
topoteretes/cognee
Cognee is an open-source project that provides scalable and modular memory capabilities for AI agents using ECL pipelines, enabling efficient management and retrieval of past interactions and diverse data types.
vectorize-io/hindsight
Hindsight is an agent memory system designed to create smarter, learning AI agents by moving beyond simple conversation history recall to encompass a biomimetic storage architecture for long-term m...
screenpipe/screenpipe
Screenpipe is a local-first, AI-powered memory tool that records computer activity (screen, audio, accessibility tree) to provide a searchable, automated context for personal AI agents and users, i...
Olow304/memvid
Memvid is a video-based AI memory library that stores millions of text chunks in MP4 files with ultra-fast semantic search, eliminating the need for traditional databases and enabling portable, efficient AI memory management.
EverMind-AI/EverOS
EverOS is an open-source Python framework for self-evolving, portable, long-term memory for AI agents, supporting multimodal ingestion and local-first persistence with Markdown as the source of truth.
MemTensor/MemOS
MemOS is a Memory Operating System for LLMs and AI agents that unifies long-term memory management, offering a single API for structured, multi-modal memory with enterprise-grade optimizations, sig...