us3r/synapse-mcp-memory-server

🧠 SYNAPSE: Self-Evolving Memory MCP Server

🧬 Where AI Memories Learn and Neural Connections Grow Stronger

  

🏆 What Makes SYNAPSE Special

SYNAPSE revolutionizes AI memory by implementing features that no other MCP server has:

• —🧬 Self-Evolving Memories: Memories adapt, strengthen, and form new connections autonomously
• —🤔 Metacognitive Awareness: The system knows what it knows and what it doesn't know
• —🕸️ Dynamic Knowledge Graphs: Zettelkasten-inspired interconnected memory networks
• —🎯 Confidence Assessment: Real-time confidence scoring for knowledge queries
• —🔍 Semantic Evolution: Memories learn from access patterns and strengthen relationships
• —📊 3D Interactive Visualization: Beautiful visualization showing memory evolution in real-time

🚀 Quick Start (2 Steps)

1. Install SYNAPSE

# Clone and install
git clone <repository-url>
cd memory_mcp
uv pip install -e .

2. Start the MCP Server

synapse serve

That's it! The server auto-initializes everything and provides:

• —MCP tools for Claude Desktop integration
• —Automatic memory storage and evolution
• —3D visualization at http://localhost:7860

🔌 Using SYNAPSE with Claude Desktop

Local Installation

Add to your Claude Desktop MCP configuration:

{
  "mcpServers": {
    "synapse": {
      "command": "synapse",
      "args": ["serve"]
    }
  }
}

Hugging Face Space (Gradio MCP)

Use SYNAPSE directly from Hugging Face Spaces:

{
  "mcpServers": {
    "synapse-gradio": {
      "url": "https://your-space-name.hf.space/gradio_api/mcp/sse"
    }
  }
}

The Gradio MCP endpoint provides the same tools with additional web interface benefits:

• —Real-time 3D visualization at the Space URL
• —Interactive memory exploration
• —Built-in demo data for immediate testing

Then use memory tools in Claude:

• —store_memory - Save information with automatic categorization
• —search_memory - Find memories using semantic similarity
• —evolve_memory - Trigger memory network evolution
• —get_memory_insights - Analyze knowledge gaps and patterns

💡 Memory Organization

SYNAPSE uses a unified memory space organized by:

# Store memories with tags and context
store_memory(
  content="Python is great for data science",
  tags=["programming", "python", "data-science"],
  context="learning",
  memory_type="semantic"
)

# Search by topic
search_memory(query="python programming", tags=["programming"])

Memory Types

• —📖 Episodic: Personal experiences ("I learned React today")
• —🧮 Semantic: Facts and concepts ("React is a JavaScript library")
• —⚙️ Procedural: How-to knowledge ("To create a React component, use...")

🎮 See SYNAPSE in Action

Want to see the features before using with Claude? Run the standalone demo:

synapse demo

This creates sample memories and shows the 3D visualization - perfect for demos and testing!

🛠️ Available Commands

synapse serve          # Start MCP server (main command)
synapse demo           # Standalone feature demo
synapse status         # Check system status  
synapse config --show  # View configuration
synapse test           # Test installation

🎨 3D Visualization

When running, visit http://localhost:7860 to see:

• —🕸️ Interactive 3D Memory Graph: Explore connections in three dimensions
• —📊 Real-time Analytics: Memory statistics and type distributions
• —🔍 Semantic Search Interface: Find and highlight relevant memories
• —🧬 Evolution Controls: Trigger and monitor memory adaptation

🧬 How Memory Evolution Works

SYNAPSE continuously evolves your memory network:

Automatic Evolution Triggers

• —Time-based: Every hour (configurable)
• —Activity-based: Every 50 new memories
• —Manual: Via MCP tools

Evolution Process

1. 1.Memory Decay: Unused memories gradually weaken
2. 2.Reinforcement: Frequently accessed memories strengthen
3. 3.Connection Discovery: New relationships form based on usage
4. 4.Network Optimization: Weak connections are pruned
5. 5.Community Detection: Related memories cluster together

🤔 Metacognitive Capabilities

SYNAPSE understands its own knowledge:

• —Confidence Scoring: How certain the system is about information
• —Knowledge Gap Detection: Identifies missing knowledge areas
• —Coverage Analysis: Maps well/poorly covered topics
• —Learning Velocity: Tracks knowledge acquisition patterns

⚙️ Configuration (Optional)

SYNAPSE works out-of-the-box, but you can customize:

# View current settings
synapse config --show

# Set custom database path
export SYNAPSE_MEMORY_DB_PATH="./my_memories/memory.db"

# Set custom visualization port  
export SYNAPSE_VIZ_PORT=8080

# Start with custom settings
synapse serve

🏗️ Architecture

SYNAPSE System
├── 🧠 Memory System
│   ├── Vector embeddings (sentence-transformers)
│   ├── Self-evolving connections
│   └── Multi-type memory support
├── 🤔 Metacognitive System  
│   ├── Knowledge gap detection
│   └── Confidence assessment
├── 🕸️ Graph Database
│   ├── NetworkX in-memory graphs
│   └── Community detection
├── 🎨 3D Visualization
│   └── Interactive Gradio interface
└── 🔌 MCP Server
    ├── Standard MCP tools/resources
    └── Memory management APIs

🏆 Why SYNAPSE Wins

Innovation

• —First self-evolving memory system in the MCP ecosystem
• —Metacognitive awareness - knows what it knows
• —Beautiful 3D visualization that actually works
• —Unified memory space with intelligent organization

Technical Excellence

• —Auto-initializing, zero-config setup
• —Scalable architecture
• —Production-ready with proper error handling
• —Clean, extensible codebase

User Experience

• —One command to get started: synapse serve
• —Immediate visual feedback via 3D interface
• —Seamless Claude Desktop integration
• —Clear, logical workflow

📊 File Structure

src/synapse/
├── cli.py           # Command-line interface
├── server.py        # Main MCP server  
├── memory.py        # Core memory system
├── metacognition.py # Metacognitive awareness
├── visualizer.py    # 3D visualization
├── graph_db.py      # Graph database
└── config.py        # Configuration

🎯 Getting Started Checklist

• —[ ] Install: uv pip install -e .
• —[ ] Start server: synapse serve
• —[ ] Add to Claude Desktop config
• —[ ] Store your first memory in Claude
• —[ ] Visit http://localhost:7860 for visualization
• —[ ] Watch your knowledge network grow and evolve!

🎬 Video Demo

📹 [Watch SYNAPSE MCP Server in Action](demo-video-placeholder.mp4)

This video demonstrates:

• —SYNAPSE MCP server integration with Claude Desktop
• —Real-time memory storage and retrieval
• —3D interactive visualization features
• —Self-evolving memory network capabilities
• —Metacognitive insights and knowledge gap analysis

🏷️ Hackathon Submission

Track: MCP Tool/Server Track Tags: mcp-tool-server-track, gradio-hackathon-2025

Built with ❤️ for the Hugging Face + Gradio Hackathon

SYNAPSE represents the future of AI memory systems - where memories learn and connections grow stronger.