The Memory Layer Your
AI Agents Deserve

MemClawz is Redis for AI agents — a shared memory system where multiple AI agents store, search, and learn from collective memories. Open source, lightning fast, built for the AI-first future.

3,774 Memories
21 Agents
5 Servers
16 Active
Get Started Star on GitHub
terminal
$ curl -X POST http://localhost:3500/api/v1/add \
-H "Content-Type: application/json" \
-d '{"content": "Deployed v2.1", "agent_id": "infraclaw"}'
✓ Memory stored • ID: mem_3775 • Score: 0.94

What is MemClawz?

A shared memory system that makes AI agent fleets smarter, faster, and more collaborative

Redis for AI Agents

Traditional AI agents operate in isolation, losing context between conversations and duplicating work across the fleet. MemClawz changes this by providing a shared memory layer where agents can store, search, and learn from collective experiences.

  • 🧠 Collective Intelligence — Agents learn from each other's experiences
  • ⚡ Lightning Fast — Sub-millisecond memory retrieval with hybrid search
  • 🔄 Auto-Enrichment — Automatic entity and temporal extraction
  • 📊 Smart Scoring — Relevance, recency, importance, and access frequency
AI Agents
🤖 Agent A 🤖 Agent B 🤖 Agent C
↓
MemClawz API
Store Search Route
↓
Memory Engine
Vector DB Neo4j Qdrant

Why MemClawz?

Built for the challenges of real AI agent deployments

⚡

Hybrid Search

BM25 keyword + vector similarity (70/30 blend) delivers the most relevant memories every time

🔄

Memory Lifecycle

8-state management: active→verified→archived→deprecated→superseded→merged→disputed→deleted

🧠

Auto-Enrichment

Automatic entity, category, and temporal extraction on every memory write

📊

Composite Scoring

Semantic similarity + recency + importance + access frequency for perfect relevance

🗜️

Smart Compaction

3-tier compaction (session, daily, weekly) with deduplication keeps memory lean

🌐

Multi-Agent Federation

HTTP push/pull protocol for sharing memories across distributed agent fleets

💤

Sleep-Time Reflection

LLM-driven pattern detection and insight generation during downtime

📋

15 Memory Types

From decisions to events to insights — structured memory classification

How It Works

Simple API, powerful capabilities

Store a Memory

curl -X POST http://localhost:3500/api/v1/add \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Deployed v2.1 to production",
    "agent_id": "infraclaw",
    "memory_type": "event"
  }'

Search Memories

curl "http://localhost:3500/api/v1/search?q=deployment+issues&limit=5"

# Returns relevant memories with scores
{
  "memories": [
    {
      "content": "Fixed nginx config issue",
      "score": 0.94,
      "agent_id": "infraclaw",
      "timestamp": "2024-03-14T15:30:00Z"
    }
  ]
}

Hybrid Search

curl "http://localhost:3500/api/v1/hybrid-search?q=port+3500+config&alpha=0.7"

# Combines keyword + vector search
# alpha=0.7 means 70% vector, 30% keyword

Agent Routing

curl "http://localhost:3500/api/v1/route?task=fix+nginx+config"

# Returns the best agent based on memory history
{
  "recommended_agent": "infraclaw",
  "confidence": 0.89,
  "reasoning": "High expertise in nginx configurations"
}

Architecture

Built on proven technologies, optimized for AI workloads

API Layer

REST API MCP Server WebSocket

Memory Engine

Hybrid Search Compaction Routing

Storage Layer

Qdrant Vector DB Neo4j Graph Redis Cache

AI Layer

Embeddings LLM Enrichment Pattern Detection

Get Started

Deploy MemClawz in under 5 minutes

Docker (Recommended)

# Start MemClawz with all dependencies
docker run -p 3500:3500 -p 6333:6333 -p 7687:7687 \
  memclawz/memclawz:latest

# Or use docker-compose
curl -O https://raw.githubusercontent.com/yoniassia/memclawz/main/docker-compose.yml
docker-compose up -d

pip install

# Install MemClawz
pip install memclawz

# Start the server
memclawz start --port 3500

# Check health
curl http://localhost:3500/health

From Source

# Clone the repository
git clone https://github.com/yoniassia/memclawz.git
cd memclawz

# Install dependencies
pip install -r requirements.txt

# Run
python -m memclawz.server

Open Source & MIT Licensed

MemClawz is completely open source and MIT licensed. We believe the future of AI is collaborative, and memory should be a shared resource, not a competitive moat.

MIT License Use anywhere, modify freely
Active Development Weekly releases, rapid iteration
Production Ready Used by multiple AI agent fleets

yoniassia/memclawz

The Redis for AI Agents - shared memory system for AI agent fleets

⭐ 1,247 stars 🍴 89 forks 🐛 12 issues
View on GitHub

Join the Community

Connect with developers building the future of AI agent systems

💬

Discord

Join our active community for support, discussions, and real-time help

Join Discord
🐛

GitHub Issues

Report bugs, request features, and contribute to the roadmap

View Issues
📚

Documentation

Comprehensive guides, API references, and tutorials

Read Docs