MemPalace

Active

Description

MemPalace is an open-source AI memory system providing a persistent long-term memory layer for AI agents, with ChromaDB vector storage and MCP protocol integration.

Key Features

Verbatim storage & semantic search — conversation history stored as-is, retrieved via semantic search without summarization or paraphrasing
Palace-style indexing — organizes memory as Wings (people/projects) → Rooms (topics) → Drawers (verbatim content)
Local-first zero API — core retrieval requires no API key, cloud, or LLM, uses ChromaDB as default local vector store
Pluggable retrieval backend — swappable via base.py interface to support alternative vector databases
Knowledge graph — temporal entity-relationship graph with validity windows, supporting add/query/invalidate/timeline
MCP Server — 29 MCP tools covering palace read/write, knowledge-graph operations, and agent diaries

Use Cases

💡 Providing persistent long-term memory for Claude Code and other AI coding assistants across sessions

💡 Storing and retrieving large conversation histories locally to protect data privacy

💡 Tracking project stakeholder relationships and decision timelines via knowledge graph

💡 Integrating memory systems into AI agent workflows through MCP tools

💡 Validating memory retrieval performance on LongMemEval and other benchmarks

Strengths & Limitations

✅ Strengths

• Actively maintained, recent updates
• High community interest (58.4k stars)
• Permissive open-source license (MIT)

Quick Start

uv tool install mempalace
mempalace init ~/projects/myapp
# Mine project file content
mempalace mine ~/projects/myapp
# Search memories
mempalace search 'why did we switch to GraphQL'
# Load context for new session
mempalace wake-up

Visit GitHub Visit Website

Related Projects

memU

14.3k · Python

Active A

A memory system for 24/7 proactive agents with MCP protocol integration, providing long-term memory management, skill storage, and proactive reasoning capabilities for continuously running AI agents.

memoryagentmcp +2

· 24/7 proactive memory agent that continuously captures and understands user intent
· File-system-like memory structure with folders, files, symlinks, and mount points for organized knowledge
· Significantly reduces LLM token cost with smaller context windows (~1/10 of comparable usage)

View full details → View on GitHub

Nocturne Memory

1.3k · Python

Active A

A lightweight, rollbackable, and visual long-term memory server for MCP agents, replacing traditional vector RAG with reliable context retention.

pythonmcpmemory +2

· MCP-based long-term memory server enabling AI to remember identity and history across sessions and models
· One soul, any engine — memory stored in independent MCP Server, not bound to any LLM, AI retains full memory when switching models
· Namespace isolation support — multiple AI personalities can have completely independent memory spaces without interference

View on GitHub

MaxKB

22.5k · Python

Active A+

MaxKB is an open-source knowledge base Q&A and agent building platform powered by LLMs, with vector retrieval, workflow orchestration, and multi-model support out of the box.

knowledge-baseragchatbot +3

· RAG Pipeline — Document upload and online crawling with automatic splitting and vectorization to reduce hallucinations
· Agentic Workflow — Powerful workflow engine, function library and MCP tool-use for complex business scenarios
· Zero-code Integration — Seamlessly integrate into third-party business systems for smart Q&A capabilities

View full details → View on GitHub

MemAgent

1.1k · Python

Stale B

A MemAgent framework that can extrapolate to 3.5M context tokens, along with a training framework for RL training of any agent workflow.

memoryagentrag +2

· Ultra-Long Context Processing: Extrapolate from 8K training context to 3.5M tokens with performance loss under 5%
· Reinforcement Learning Driven: Trained with RLVR (Reinforcement Learning from Verifiable Rewards), extends DAPO algorithm for end-to-end multi-turn conversation optimization
· Linear Time Complexity: Breaks through computational bottlenecks in long-text processing with linear resource scaling