Memstate AI Demos

This repository contains example code and tutorials for building memory-powered AI agents using Memstate AI.

LangChain & LangGraph Agents

The langchain-agents directory contains three demos showing how to integrate Memstate with LangChain and LangGraph:

1. Personal Assistant: A simple ReAct agent that remembers user preferences across multiple sessions.
2. Multi-Agent Research: A LangGraph swarm of specialized agents (Researcher, Analyst, Writer) that share a single Memstate project to collaborate on a task.
3. Versioned Knowledge Base: A demonstration of Memstate's version control and time-travel capabilities for auditing agent memory.

Getting Started

1. Get a free API key at memstate.ai/dashboard.
2. Clone this repository.
3. Navigate to the specific demo directory and follow its README instructions.

What is Memstate?

Memstate gives your AI agents structured, versioned memory they can navigate. Custom LLM models extract keypaths automatically, detect conflicts, and compress context.

• Keypaths: Memories are organized hierarchically using dot-separated paths like auth.provider.
• Version Control: Every memory is versioned. When content is updated, the previous version is preserved in history.
• Token Efficiency: Structured keypath = value atoms instead of text blobs.
• Time-Travel Queries: See memory state at any point in history.

Learn more in the Memstate Documentation.