Introduction to LimbicDB

LimbicDB is an industrial-grade AI memory infrastructure designed to give generative agents a cognitive memory graph. Instead of relying on raw vector databases that store disjointed snippets of text, LimbicDB processes and structures information simulating human cognitive decay (ACT-R theory), embedding contextual connections, and preserving safety with enterprise-grade encryption.

Quick Start (5 lines of code)

import { open } from 'limbicdb'

const db = open('./agent.limbic')       // Create or open a memory database
await db.remember('User prefers dark mode and speaks Chinese')
const result = await db.recall('What are the user preferences?')
console.log(result.memories)            // → [{ content: 'User prefers...', strength: 0.5, ... }]
await db.close()

That's it. No configuration, no API keys, no external services. Everything runs locally.

What makes it unique?

LimbicDB isn't another vector wrapper. It is a purpose-built memory engine.

• 🧠 Cognitive Memory Graph: Memories are stored as nodes with temporal and semantic edges. The system understands what happened first, what happened often, and what happened together.
• 📉 ACT-R Decay Model: Built-in forgetting curves. Not all memories are equal. Important memories resurface, while noise gently decays over time.
• 🔐 End-to-End Encryption: Memories are encrypted using AES-256-GCM. The actual vector embeddings are decoupled from the sensitive content payloads.
• 🔌 Pluggable Ecosystem: Support for SQLite for local embedded systems, PostgreSQL (via pgvector) for cloud deployments, and direct compatibility with LangChain and LlamaIndex.
• 💻 LimbicDB Studio: A powerful desktop application (.exe, .dmg) giving you X-Ray vision into your agent's mind in real-time.
• 📡 Real-Time SSE Streaming: REST API with Server-Sent Events — any client can subscribe to live memory changes.

How it compares

Feature	LimbicDB	ChromaDB	Pinecone	Mem0
Local-first (offline)	✅	✅	❌ (cloud)	✅
Cognitive decay (ACT-R)	✅	❌	❌	❌
Memory graph (edges/links)	✅	❌	❌	Partial
E2E encryption	✅ (AES-256-GCM)	❌	❌	❌
Visual Studio (desktop app)	✅	❌	❌	❌
MCP server (AI tool)	✅	❌	❌	✅
Python + TypeScript	✅	✅	✅	✅
PostgreSQL backend	✅ (pgvector)	❌	N/A	✅
Consolidation (auto-dedup)	✅	❌	❌	❌

Architecture Overview

graph TB
    subgraph Clients
        A[Your AI Agent]
        B[MCP Server]
        C[REST API]
        D[CLI]
        E[Studio Desktop]
    end

    subgraph Core Engine
        F[LimbicDB Core]
        G[ACT-R Decay Engine]
        H[Memory Graph]
        I[Consolidation Engine]
        J[Insight Engine]
    end

    subgraph Storage
        K[SQLite + FTS5]
        L[PostgreSQL + pgvector]
        M[In-Memory]
    end

    subgraph Security
        N[AES-256-GCM Vault]
        O[Schema Validation]
    end

    A --> F
    B --> F
    C --> F
    D --> F
    E --> F
    F --> G
    F --> H
    F --> I
    F --> J
    F --> K
    F --> L
    F --> M
    F --> N
    F --> O

Installation

Core SDK (TypeScript / Node.js)

npm install limbicdb

Python SDK

pip install limbicdb

CLI Support

npm install -g limbicdb
limbic remember "I like to drink coffee in the morning."

Next Steps

• Explore the Architecture in depth
• Learn about Integrations (LangChain, LlamaIndex, MCP)
• Discover how to Secure your Memories
• Check the Roadmap for upcoming features