MemVaultDocumentation

Welcome to MemVault

MemVault is a Long-Term Memory as a Service platform designed specifically for AI agents. Built on GraphRAG technology, it enables your AI systems to store, retrieve, and intelligently consolidate memories over time.

What is MemVault?

Traditional AI models are stateless. They forget conversations the moment they end. MemVault solves this by providing a persistent memory layer that your AI agents can access across sessions, deployments, and even different applications.

Unlike simple vector databases, MemVault uses a hybrid approach that combines:

  • Vector similarity search for semantic understanding
  • Knowledge graph traversal for relationship-aware retrieval
  • Automatic consolidation through intelligent sleep cycles

This creates a memory system that doesn't just store information—it organizes, refines, and improves over time.

Key Features

Hybrid Search

MemVault combines two powerful search paradigms:

Vector Similarity: Every memory is converted into a high-dimensional embedding using state-of-the-art models. This allows semantic search where "password reset" and "forgot my login" are understood as related concepts.

Knowledge Graph Traversal: Memories aren't isolated—they're connected through a graph structure. When you retrieve a memory, MemVault can traverse relationships to find contextually relevant information that pure vector search might miss.

The result? More accurate, context-aware memory retrieval that understands both meaning and relationships.

Self-Organizing Sleep Cycles

This is MemVault's core differentiator.

Just like the human brain consolidates memories during sleep, MemVault runs automated Sleep Cycles every 6 hours. During these cycles, the system:

  • Merges duplicate or highly similar memories to reduce redundancy
  • Strengthens important connections between frequently co-accessed memories
  • Archives low-value memories to keep your active memory space efficient
  • Updates relationship weights in the knowledge graph based on usage patterns
  • Cleans up noise and irrelevant data automatically

You don't need to manually manage memory hygiene—the system does it for you.

Note: Sleep Cycles are available on Direct and Pro plans. Hobby users get standard storage without automatic consolidation.

Platform Agnostic

MemVault works with any LLM or AI framework:

  • OpenAI GPT models (GPT-4, GPT-3.5)
  • Anthropic Claude (Claude 3, Claude 2)
  • Open-source models (Llama, Mistral, etc.)
  • Custom AI systems built with LangChain, LlamaIndex, or from scratch

Use our official TypeScript/JavaScript SDK or call the REST API directly from any language.

Official SDK Available

Get started faster with our official SDK:

npm install @memvault/client

NPM Package: @memvault/client

import { MemVault } from '@memvault/client';

const memvault = new MemVault(process.env.MEMVAULT_API_KEY);

// Store a memory
await memvault.addMemory("Meeting was about Q4 budget");

// Search memories
const results = await memvault.retrieve("budget decisions");
console.log(results.memories);

Learn more in the SDK Quick Start Guide.

Architecture Overview

┌─────────────────┐
│   Your AI App   │
└────────┬────────┘
         │ HTTP API
         ▼
┌─────────────────┐
│   MemVault API  │
│  (Koyeb Worker) │
└────────┬────────┘
         │
    ┌────┴────┐
    ▼         ▼
┌────────┐  ┌──────────────┐
│ pgvector│  │ Redis Queue  │
│PostgreSQL  │  (BullMQ)    │
└─────────┘  └──────┬───────┘
                    │
             ┌──────▼───────┐
             │ Sleep Cycle  │
             │  Processor   │
             └──────────────┘

Use Cases

AI Chatbots & Assistants

Give your chatbot long-term memory so it remembers user preferences, past conversations, and important context across sessions.

Content Generation Systems

Maintain consistency in generated content by retrieving relevant past outputs and user feedback.

Customer Support Automation

Remember customer issues, preferences, and history to provide personalized support experiences.

Knowledge Management

Build a searchable knowledge base that understands semantic relationships between documents and concepts.

CI/CD Intelligence

Store deployment history, test results, and incident reports to help your CI/CD pipelines learn from past events.

Getting Started

Ready to give your AI long-term memory? Here's what to do next:

  1. SDK Quick Start - Install the official SDK and get running in 5 minutes
  2. Authentication - Learn how to secure your API requests
  3. GitHub Actions Integration - Automate memory storage in CI/CD
  4. Architecture Deep Dive - Understand how Sleep Cycles work

Pricing & Plans

MemVault offers three tiers:

  • Hobby ($29/month): Standard storage and retrieval, no sleep cycles
  • Direct ($99/month): Includes sleep cycles, priority support, higher rate limits
  • Pro ($299/month): Enterprise features, custom consolidation rules, dedicated support

All plans include unlimited API keys, GraphRAG retrieval, and PostgreSQL + pgvector storage.

Questions? Check out our SDK Documentation or contact support at support@memvault.com.