MEMOROUTER MEMOROUTER

Persistent Identity Agents: Overcoming LLM Amnesia

The next generation of sovereign artificial intelligence. Agents that never forget who they are, evolve continuously with human guidance, consolidate knowledge in the Sleep Module, operate in Docker Sandboxes, and stream neural voice to devices for a natural voice conversation.

Explore Persistent Identity ↓
memorouter:core-orchestrator // session_state.active
Multi-Agent Isolation: Active
// 1. Multimodal Perception & Sovereign Interlocutor Identification
INBOUND [Channel: WhatsApp / LID: 5511999998888] → User: "Alex Mercer" (Authenticated Owner)
RAG_HYBRID_RRF → pgvector 384d (0.942) + FTS Trigram (0.890) → RRF Top Score: 0.032 [Context: "Prefers modular sandboxes & E2EE"]
ENTITY_GRAPH → 2-Hop Traversal: (Alex Mercer → owns → CoreEngine → requires → ZeroLockIn)
// 2. Deep Research Engine & Official Knowledge Base
DEEP_RESEARCH → Recursive Synthesis [Sub-topics: 3] → Auto-Citation evidence verified: 100% match
// 3. Sleep Module & Continuous Self-Learning
SLEEP_ENGINE → Nightly consolidation complete: 18 procedural connections synthesized, memory decay applied
// 4. Docker Sandbox Execution & Tool Dispatching
EXEC_SANDBOX → Docker Container [task_sandbox_01] → Running tests in isolated environment... OK (0 failures)
// 5. Real-time Audio Stream to ESP32 Hardware & Messaging
VOICE_STREAM → BinaryProtocol3 Opus frame stream → ESP32-S3 Speaker [MAC: 34:85:18:a2:bc:10] (16kHz / 24kbps)
// 6. Two-Tier Skills Engine & Biological Hebbian Reinforcement
SKILLS_ROUTER → Context-Aware Two-Tier Prompt [Tier 1: deploy-release / Tier 2: 14 indexed] → 75% prompt tokens saved
HEBBIAN_SYNC → Co-occurrence reinforced: (memo → applies_to → deploy-release: +0.05 confidence → 0.95)
THE AUTONOMOUS AGENT REVOLUTION

Why Persistent Identity Agents Are the New Frontier of AI

Standard Large Language Models are stateless (suffering from digital amnesia after each session). MEMOROUTER transforms LLMs into continuous cognitive entities with cumulative memory, immutable identity anchors, and user-guided behavioral evolution.

βš“

Identity Anchors (Deterministic Core)

The agent core values, tone, immutable directives, and personality reside in deterministic identity anchors loaded at every turn. The agent never suffers from identity drift, even across months of continuous operation.

0% Personality Drift // Immutable Values
🧠

Tripartite Memory (Episodic, Semantic & Procedural)

Instead of an unorganized message buffer, memory is structured like the human brain: Episodic (past experiences & timestamps), Semantic (user facts & preferences), and Procedural (practical skills & tool mastery).

Human-Grade Cognitive Hierarchy
πŸŒ™

Sleep Module & Continuous Self-Learning

During periodic sleep routines, the agent processes daily interactions, extracts user corrections, synthesizes new permanent procedural rules, and discards ephemeral noise to maintain a compact, cost-effective context window.

Automated Knowledge Synthesis
Smart Fallback
Intelligent API fallback & OpenRouter routing support
100%
Real-time voice pipeline and continuous streaming
Nightly
Nightly consolidation of long-term memories
Docker
Secure execution of code and automations
FEATURES

High-Precision Cognitive Engineering Features

Built for end-to-end production: from embedded microcontrollers to orchestration servers.

🧠 Hybrid RAG

True Hybrid RRF RAG & Vector Memory

Fusion of dense vector search (pgvector 384d/1536d) and trigram/FTS full-text search via Reciprocal Rank Fusion with biological decay and dynamic contextual scoring.

pgvector // 384d & 1536d Hybrid RRF
πŸ”¬ Deep Research

Deep Research & Official Knowledge Base

Autonomous deep research engine featuring recursive sub-topic planning, multi-source synthesis, hierarchical indexing, and strict evidence-based citation.

deep_research // Recursive Synthesis
πŸ•ΈοΈ Graph RAG & Skills

Entity Knowledge Graph & Two-Tier Skills Engine

Autonomous extraction of named entities, RDF relation triples, and bidirectional skill linking with two-tier context-aware prompt injection and Hebbian usage reinforcement.

entity_graph // 2-Hop Traversal & Skills Engine
πŸŒ™ Sleep Engine

Sleep Engine & Memory Consolidation

Biological cognitive cycle that cleans daily buffers, prunes low-importance memories, synthesizes episodic experiences into permanent procedural rules, and re-indexes vectors.

sleep_routine // Memory Consolidation
πŸ“¦ Sandbox

Docker Sandboxes & Autonomous CLI

Spin up isolated Docker containers for secure command execution, code compilation, automated test suites, and direct Git repository management.

mcp_docker // Isolated Execution
πŸŽ™οΈ Omnichannel

Omnichannel & Real-time Neural Voice

Simultaneous messaging on WhatsApp, Telegram, Discord, and Slack, paired with low-latency binary Opus streaming for ESP32 microcontrollers and Kokoro/Piper neural TTS.

binary_opus // Kokoro Neural Voice
πŸ”€ Multi-LLM

Multi-LLM Routing & Smart Fallback

Native connectivity to OpenRouter, NVIDIA NIM, DeepSeek, Google Gemini, Anthropic Claude, and OpenAI, featuring dynamic fault-tolerant fallback switching.

openrouter_routing // 300+ Models
πŸ›‘οΈ E2EE Sovereign

Sovereign E2EE Tunnel & MEMO Desktop

Native desktop runtime featuring local PostgreSQL, pgvector, and a Zero-Knowledge End-to-End Encrypted tunnel (X25519 + AES-256-GCM) for absolute data sovereignty.

memo_desktop // X25519 E2EE
LOCAL RUNTIME & VECTOR MEMORY

MEMO Desktop: Sovereign Local Memory & Gateway

Run your dedicated vector database with PostgreSQL pgvector, local embeddings via Ollama, and seamless WhatsApp bridge directly on your machine without relying on third-party memory clouds.

πŸŒ‰ Sovereign Local Architecture

MEMO Desktop is a standalone desktop application that runs in the system tray, providing a local PostgreSQL database with pgvector, Ollama embeddings, WhatsApp integration, and Traccar GPS fleet/vehicle tracking. It enables your agents to store memories, monitor assets, and automate routines with complete data sovereignty and zero cloud lock-in.

  • πŸ”’ End-to-End Encryption (E2EE): Zero-Knowledge tunnel with X25519 & AES-256-GCM
  • πŸ›‘οΈ Cryptographic Safety Number verification on both desktop and dashboard
  • βœ“ PostgreSQL 16 + pgvector for fast semantic search
  • βœ“ Local Ollama inference for embeddings & private processing
  • βœ“ Native WhatsApp bridge & real-time webhook routing
  • βœ“ One-click installer with automatic background service
Download MEMO Desktop ⬇️
MEMO Desktop
Service Active
MEMO Desktop
ENGINEERED FOR EXTREME AUTONOMY

Built for Visionaries, Engineers & Enterprise Sovereignty

Whether you are automating your personal digital life, scaling multi-agent software engineering, or demanding an enterprise alternative to Big Tech cloud lock-in β€” MEMOROUTER delivers uncompromised cognitive autonomy.

⚑
ASSISTANT & IOT

AI Enthusiasts & Smart Hardware

Ambient Intelligence β€’ True LTM β€’ ESP32 Neural Voice

  • 🧠
    Continuous Tripartite Memory: An assistant that never forgets your habits, preferences, and projects across months without re-prompting.
  • πŸŽ™οΈ
    Sub-Second Binary Opus Voice: Ultra-low latency streaming directly to ESP32-S3 hardware, desktop speakerphones, and WhatsApp voice notes.
  • πŸ›°οΈ
    Real-Time GPS & IoT Telemetry: Integrated Traccar GPS fleet/vehicle tracking, desktop PC telemetry, and local smart device actuation.
  • πŸ›‘οΈ
    100% Private & Sovereign: Memories reside in your local PostgreSQL pgvector database, protected by Zero-Knowledge E2EE (X25519).
hardware_ready // ESP32 + Traccar GPS
πŸ’»
DEVS & SOFTWARE HOUSES

Developers & AI Engineers

Multi-Agent Swarms β€’ Two-Tier Skills β€’ FastMCP Arsenal

  • πŸ“¦
    Autonomous Docker Sandboxes: Spin up disposable Linux containers per agent with workspace mounts for compiling, testing, and git operations.
  • ⚑
    Two-Tier Skills & Hebbian Learning: Cuts system prompt token overhead by up to 75% via context-aware skill injection and biological usage reinforcement.
  • πŸ”§
    Comprehensive FastMCP Arsenal: Chrome DevTools browser automation, repository filesystem tools, deep research engine, and encrypted credentials vault.
  • πŸ”€
    Smart Multi-LLM Router (80% Savings): Route code to DeepSeek, Claude 3.7 Sonnet, GPT-4.5, or local Ollama with automatic failover and zero vendor lock-in.
multi_agent // Docker + FastMCP Tools
πŸ›οΈ
INVESTORS & BIG TECH

Why Big Tech Is Watching

Edge/Cloud Paradigm β€’ Offline Sleep Moat β€’ Zero-Knowledge

  • 🌐
    The Edge/Cloud Cognitive Paradigm: Big Tech locks enterprise data into centralized recurring silos. We decouple intelligence routing from sovereign edge storage.
  • πŸŒ™
    Sleep Engine vs. Context Waste: Instead of brute-forcing costly 1M-token prompt windows, our bio-inspired sleep engine compiles 2-hop graphs offline, cutting inference costs by 10x.
  • πŸ”’
    Zero-Knowledge Enterprise Moat: Hardware-grade E2EE (X25519 + AES-256-GCM) unlocks enterprise adoption (HIPAA, LGPD, banking secrets) where cloud AI is legally banned.
  • βš–οΈ
    Absolute Vendor Independence: Completely immune to single-vendor outages, pricing hikes, or censorship policies across 300+ frontier and local LLMs.
sovereign_moat // Zero Cloud Lock-in
INTEGRATED APIS & MODELS

Native Support for Industry-Leading LLM APIs

Bring your own API keys or connect local inference servers. Full freedom of choice per agent with zero vendor lock-in.

OpenRouter 300+ AI Models
NVIDIA NIM NeMo & Llama-Nemotron
DeepSeek V3 & R1 Reasoner
Google Gemini Live & 2.0 Flash/Pro
Anthropic Claude 3.5 Sonnet
OpenAI GPT-4o & o1/o3
Groq Ultra-low Latency LPU
xAI Grok Grok-2 & Vision
Ollama 100% Offline LLMs
Hume AI Empathic Voice
OMNICHANNEL & IOT

Supported Ecosystem & Providers

Full interoperability with embedded hardware, instant messengers and frontier LLMs.

πŸ“± WhatsApp ✈️ Telegram Bots πŸ‘Ύ Discord Channels πŸ’Ό Slack Workspaces πŸ€– Hermes πŸ—£οΈ ElevenLabs πŸŽ™οΈ Hume AI Voice ✨ Google Gemini 🧠 OpenAI GPT-4o πŸ¦™ Ollama Local LLMs

Ready to orchestrate your sovereign cognitive AI?

Access the cognitive console, configure your agents and bind your devices in minutes.