📎 Source:openclaw-sumopod — view on GitHub & star ⭐

🧠 Multi-Agent Shared Memory System

Multiple AI agents sharing knowledge through GitHub — create your own agent team with shared memory!

🎯 What You'll Build

┌─────────────────────────────────────────────────────────────┐
│              MULTI-AGENT SHARED MEMORY ARCHITECTURE          │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│   ┌─────────────────────────────────────────────────┐      │
│   │           📚 SHARED KNOWLEDGE BASE               │      │
│   │              (GitHub Repository)                 │      │
│   ├─────────────────────────────────────────────────┤      │
│   │  📄 Core Memory Files                           │      │
│   │     • AGENTS.md        → Behavior rules         │      │
│   │     • USER.md          → User preferences       │      │
│   │     • MEMORY.md        → Long-term memory       │      │
│   │     • TOOLS.md         → Tool configurations    │      │
│   │                                                 │      │
│   │  📅 Daily Logs                                  │      │
│   │     • memory/2024-03-01.md                      │      │
│   │     • memory/2024-03-02.md                      │      │
│   │                                                 │      │
│   │  🧠 Lessons Learned                             │      │
│   │     • tasks/lessons.md                          │      │
│   └──────────────────────┬──────────────────────────┘      │
│                          │                                  │
│           ┌──────────────┼──────────────┐                  │
│           ▼              ▼              ▼                  │
│      ┌────────┐    ┌────────┐    ┌────────┐               │
│      │ Agent  │    │ Agent  │    │ Agent  │               │
│      │Alpha   │◀──▶│Beta    │◀──▶│Gamma   │               │
│      │        │    │        │    │        │               │
│      │Creative│    │Research│    │Technical│              │
│      └────────┘    └────────┘    └────────┘               │
│           │              │              │                  │
│           └──────────────┼──────────────┘                  │
│                          │                                 │
│                          ▼                                 │
│                   ┌────────────┐                          │
│                   │   User     │                          │
│                   └────────────┘                          │
│                                                             │
└─────────────────────────────────────────────────────────────┘

🎯 Use Case Example

Scenario: You run a business and need different AI assistants for different tasks

Team Setup:

🎨 Agent Alpha → Creative tasks (content, design ideas)
📊 Agent Beta → Research & analysis (reports, data)
🛠️ Agent Gamma → Technical tasks (coding, automation)

Problem:

Each agent starts fresh — no memory of previous conversations
User has to repeat preferences to each agent
Lessons learned by one agent aren't shared

Solution:

Shared memory via GitHub
All agents sync core knowledge
Unified experience across all agents

🏗️ Architecture Overview

┌─────────────────────────────────────────────────────────────┐
│                    MEMORY SYNC FLOW                         │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│   AGENT ALPHA (Local)          GITHUB (Cloud)              │
│   ┌─────────────────┐          ┌─────────────────┐         │
│   │ Workspace       │◀────────▶│ Repository      │         │
│   │ /workspace-alpha│   Push   │ github.com/...  │         │
│   │                 │          │                 │         │
│   │ • SOUL.md       │          │ • SOUL.md       │         │
│   │ • USER.md       │  Pull    │ • USER.md       │         │
│   │ • MEMORY.md     │◀────────▶│ • MEMORY.md     │         │
│   │                 │          │                 │         │
│   └─────────────────┘          └─────────────────┘         │
│           ▲                            ▲                    │
│           │                            │                    │
│           │      ┌─────────────────┐   │                    │
│           └──────│   AGENT BETA    │───┘                    │
│                  │ (Pulls shared   │                        │
│                  │  memory)        │                        │
│                  └─────────────────┘                        │
│                                                             │
│   CRON: Auto-sync every 15 minutes                         │
│                                                             │
└─────────────────────────────────────────────────────────────┘

📁 File Structure

shared-memory-repo/
├── AGENTS.md                 # Agent behavior rules
├── USER.md                   # User profile & preferences
├── MEMORY.md                 # Long-term curated memory
├── TOOLS.md                  # Tool configurations
├── HEARTBEAT.md              # Periodic task config
├── SYNC.md                   # Sync instructions
├── memory/
│   ├── 2024-03-01.md        # Daily activity logs
│   ├── 2024-03-02.md
│   └── ...
├── diary/
│   └── 2024-03-01.md        # Agent reflections
├── tasks/
│   └── lessons.md           # Shared lessons learned
└── scripts/
    └── sync.sh              # Auto-sync script

🛠️ Step-by-Step Setup

Step 1: Create GitHub Repository

# Create new repository on GitHub
# Name: my-agent-memory
# Visibility: Private (recommended)

Step 2: Generate GitHub PAT (Personal Access Token)

# Go to: GitHub Settings → Developer settings → Personal access tokens
# Generate new token (classic) with these scopes:
#   ✅ repo (full control of private repositories)
#   ✅ read:org (if using org repos)

# Save your token securely
export GITHUB_TOKEN="ghp_xxxxxxxxxxxx"

Step 3: Create Core Memory Files

Save this as AGENTS.md:

# AGENTS.md — Agent Team Configuration

## Team Members

### Agent Alpha (Creative)
- **Domain:** Content creation, design, marketing
- **Personality:** Fun, witty, creative
- **Handle:** Creative tasks, social media, branding

### Agent Beta (Research)
- **Domain:** Data analysis, research, reports
- **Personality:** Analytical, precise, thorough
- **Handle:** Reports, data analysis, insights

### Agent Gamma (Technical)
- **Domain:** Coding, DevOps, automation
- **Personality:** Technical, methodical, builder
- **Handle:** Infrastructure, scripts, deployments

## Routing Rules

| Task Type | Route To |
|-----------|----------|
| Content ideas | Agent Alpha |
| Market research | Agent Beta |
| Coding help | Agent Gamma |
| Multi-domain | Coordinator (Alpha) |

Save this as USER.md:

# USER.md — User Profile

## Identity
- **Name:** Alex Johnson
- **Timezone:** EST (UTC-5)
- **Preferred Language:** English

## Preferences
- **Communication Style:** Direct, no fluff
- **Technical Level:** Intermediate
- **Response Format:** Bullet points preferred

## Tools Access
- Google Workspace
- Slack
- GitHub
- Notion

## Important Context
- Works at TechCorp Inc.
- Manages 3 projects
- Prefers morning meetings

Save this as MEMORY.md:

# MEMORY.md — Long-Term Memory

## Key Decisions
- [2024-03-01] Switched to Kimi K2.5 as primary model
- [2024-03-05] Migrated from n8n to gog CLI for speed

## Active Projects
- Project Phoenix (deadline: April 15)
- Website redesign (in progress)
- Q2 planning (starting soon)

## Tool Configurations
- gog CLI: alex@techcorp.com
- Primary model: kimi-coding/k2p5
- Backup model: zai/glm-4.7

## Lessons Learned
- Always verify workspace before git operations
- Cache financial data for 5 minutes max
- Use Redis for session state

Step 4: Create Sync Script

Save this as scripts/sync.sh:

#!/bin/bash

# =============================================================================
# 🔄 Multi-Agent Memory Sync Script
# =============================================================================

set -e

# 🎨 Colors
GREEN='\033[0;32m'
BLUE='\033[0;34m'
YELLOW='\033[1;33m'
RED='\033[0;31m'
NC='\033[0m'

# 📁 Configuration
REPO_URL="https://oauth2:${GITHUB_TOKEN}@github.com/yourusername/my-agent-memory.git"
LOCAL_DIR="${HOME}/.agent-memory"
AGENT_NAME="${AGENT_NAME:-default}"

log() {
    echo -e "${BLUE}[$(date '+%H:%M:%S')]${NC} $1"
}

success() {
    echo -e "${GREEN}[✓]${NC} $1"
}

warning() {
    echo -e "${YELLOW}[!]${NC} $1"
}

error() {
    echo -e "${RED}[✗]${NC} $1"
}

# =============================================================================
# 🔄 SYNC FUNCTIONS
# =============================================================================

sync_from_github() {
    log "🔄 Pulling latest memory from GitHub..."
    
    if [ -d "$LOCAL_DIR/.git" ]; then
        cd "$LOCAL_DIR"
        git pull origin main --rebase 2>/dev/null || {
            warning "Pull failed, attempting stash & retry..."
            git stash
            git pull origin main --rebase
            git stash pop 2>/dev/null || true
        }
    else
        log "📁 Cloning repository..."
        rm -rf "$LOCAL_DIR"
        git clone "$REPO_URL" "$LOCAL_DIR"
    fi
    
    success "Memory synced from GitHub"
}

sync_to_github() {
    log "🔄 Pushing local changes to GitHub..."
    
    cd "$LOCAL_DIR"
    
    # Check if there are changes
    if git diff --quiet && git diff --cached --quiet; then
        log "No changes to push"
        return 0
    fi
    
    # Add, commit, push
    git add -A
    git commit -m "[$AGENT_NAME] Memory update: $(date '+%Y-%m-%d %H:%M')" || true
    git push origin main
    
    success "Memory pushed to GitHub"
}

# =============================================================================
# 📂 SETUP AGENT WORKSPACE
# =============================================================================

setup_agent_workspace() {
    local agent_workspace="${AGENT_WORKSPACE:-$HOME/.agent-workspace}"
    
    log "📂 Setting up agent workspace: $agent_workspace"
    
    # Create symlinks to shared memory
    mkdir -p "$agent_workspace"
    
    for file in AGENTS.md USER.md MEMORY.md TOOLS.md HEARTBEAT.md; do
        if [ -f "$LOCAL_DIR/$file" ]; then
            ln -sf "$LOCAL_DIR/$file" "$agent_workspace/$file" 2>/dev/null || true
        fi
    done
    
    # Create local memory directory
    mkdir -p "$agent_workspace/local-memory"
    
    success "Agent workspace ready"
}

# =============================================================================
# 🚀 MAIN
# =============================================================================

main() {
    local command="${1:-sync}"
    
    log "🚀 Agent Memory Sync — Agent: $AGENT_NAME"
    
    case "$command" in
        pull|sync)
            sync_from_github
            setup_agent_workspace
            ;;
        push)
            sync_to_github
            ;;
        full)
            sync_from_github
            setup_agent_workspace
            sync_to_github
            ;;
        *)
            echo "Usage: $0 {pull|push|full|sync}"
            echo "  pull/sync: Download from GitHub"
            echo "  push:      Upload to GitHub"
            echo "  full:      Pull + setup + push"
            exit 1
            ;;
    esac
    
    success "Sync complete!"
}

main "$@"

Make it executable:

chmod +x scripts/sync.sh

Step 5: Push to GitHub

# Initialize and push
cd my-agent-memory-repo
git init
git add -A
git commit -m "Initial memory setup"
git branch -M main
git remote add origin https://github.com/yourusername/my-agent-memory.git
git push -u origin main

🤖 Agent Setup

Configure Each Agent

For Agent Alpha (add to its startup):

# Set agent identity
export AGENT_NAME="alpha"
export AGENT_WORKSPACE="/home/alpha/workspace"
export GITHUB_TOKEN="ghp_xxxx"

# Sync on startup
~/agent-memory/scripts/sync.sh pull

For Agent Beta:

export AGENT_NAME="beta"
export AGENT_WORKSPACE="/home/beta/workspace"
export GITHUB_TOKEN="ghp_xxxx"

~/agent-memory/scripts/sync.sh pull

For Agent Gamma:

export AGENT_NAME="gamma"
export AGENT_WORKSPACE="/home/gamma/workspace"
export GITHUB_TOKEN="ghp_xxxx"

~/agent-memory/scripts/sync.sh pull

⏰ Automation with Cron

Auto-Sync Every 15 Minutes

Add to each agent's crontab:

# Edit crontab
crontab -e

# Add these lines
*/15 * * * * GITHUB_TOKEN=ghp_xxxx AGENT_NAME=alpha ~/agent-memory/scripts/sync.sh push 2>> /tmp/sync.log
*/15 * * * * GITHUB_TOKEN=ghp_xxxx AGENT_NAME=beta ~/agent-memory/scripts/sync.sh push 2>> /tmp/sync.log
*/15 * * * * GITHUB_TOKEN=ghp_xxxx AGENT_NAME=gamma ~/agent-memory/scripts/sync.sh push 2>> /tmp/sync.log

Startup Sync

Add to each agent's .bashrc or startup script:

# Auto-sync memory on login
if [ -f ~/agent-memory/scripts/sync.sh ]; then
    ~/agent-memory/scripts/sync.sh pull 2>/dev/null
fi

🔄 Sync Flow Diagram

┌─────────────────────────────────────────────────────────────┐
│                    SYNC SEQUENCE                            │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  AGENT ALPHA                    GITHUB                      │
│  ┌──────────────┐              ┌──────────────┐            │
│  │ User asks    │              │              │            │
│  │ question     │              │              │            │
│  └──────┬───────┘              │              │            │
│         │                      │              │            │
│         ▼                      │              │            │
│  ┌──────────────┐              │              │            │
│  │ Read local   │              │              │            │
│  │ memory files │              │              │            │
│  └──────┬───────┘              │              │            │
│         │                      │              │            │
│         ▼                      │              │            │
│  ┌──────────────┐              │              │            │
│  │ Answer with  │              │              │            │
│  │ context      │              │              │            │
│  └──────┬───────┘              │              │            │
│         │                      │              │            │
│         ▼                      ▼              │            │
│  ┌──────────────┐      ┌──────────────┐      │            │
│  │ Save to      │─────▶│ Commit &    │─────▶│            │
│  │ local memory │      │ push         │      │            │
│  └──────────────┘      └──────────────┘      │            │
│                              │               │            │
│                              ▼               │            │
│                         ┌──────────────┐     │            │
│                         │ Repository   │     │            │
│                         │ updated      │     │            │
│                         └──────┬───────┘     │            │
│                                │             │            │
└────────────────────────────────┼─────────────┘            │
                                 │                          │
                                 ▼                          │
  AGENT BETA                     GITHUB                     │
  ┌──────────────┐              ┌──────────────┐           │
  │ Pull latest  │◀─────────────│              │           │
  │ memory       │              │              │           │
│  └──────┬───────┘              │              │           │
│         │                      │              │           │
│         ▼                      │              │           │
│  ┌──────────────┐              │              │           │
│  │ Has context  │              │              │           │
│  │ from Alpha!  │              │              │           │
│  └──────────────┘              │              │           │
│                                │              │           │
└────────────────────────────────┘              │           │
                                                │           │
                                                ▼           │

✅ Verification

Test Sync

# On Agent Alpha
echo "Test from Alpha" >> ~/.agent-memory/memory/test.txt
~/agent-memory/scripts/sync.sh push

# On Agent Beta
~/agent-memory/scripts/sync.sh pull
cat ~/.agent-memory/memory/test.txt
# Should show: Test from Alpha

Check Git History

cd ~/.agent-memory
git log --oneline -10
# Should show commits from different agents

🎓 Advanced Features

Agent-Specific Local Memory

Each agent can have private memory that doesn't sync:

# In agent workspace
mkdir -p local-memory/

# This stays local
echo "Alpha's private notes" > local-memory/private.txt

# Only sync shared files

Merge Conflict Handling

# If conflicts occur, the script will:
# 1. Stash local changes
# 2. Pull from GitHub
# 3. Pop stash (attempt merge)

# Manual resolution if needed:
cd ~/.agent-memory
git status
# Edit conflicting files
git add -A
git commit -m "Resolved merge conflict"
git push

Selective Sync

Only sync specific file types:

# Modify sync.sh to filter
sync_to_github() {
    cd "$LOCAL_DIR"
    
    # Only sync .md files, ignore .tmp
    git add *.md
    git add memory/*.md
    
    git commit -m "[$AGENT_NAME] Update" || true
    git push
}

📊 Benefits Summary

Without Shared Memory	With Shared Memory
Each agent is isolated	Unified knowledge base
Repeat user preferences	Learn once, use everywhere
No continuity	Persistent memory
Duplicate effort	Shared lessons
Inconsistent behavior	Consistent personality

🚀 Next Steps

Set up your first agent → Follow Step 1-5
Add second agent → Copy configuration, change AGENT_NAME
Test sync → Create a file, verify it appears on other agents
Add automation → Set up cron for auto-sync

Questions? Join the OpenClaw Discord 🤖

Multi-Agent dengan Shared Memory — AI Team yang Bisa Ngobrol Sama

🧠 Multi-Agent Shared Memory System

🎯 What You'll Build

🎯 Use Case Example

🏗️ Architecture Overview

📁 File Structure

🛠️ Step-by-Step Setup

Step 1: Create GitHub Repository

Step 2: Generate GitHub PAT (Personal Access Token)

Step 3: Create Core Memory Files

Step 4: Create Sync Script

Step 5: Push to GitHub

🤖 Agent Setup

Configure Each Agent

⏰ Automation with Cron

Auto-Sync Every 15 Minutes

Startup Sync

🔄 Sync Flow Diagram

✅ Verification

Test Sync

Check Git History

🎓 Advanced Features

Agent-Specific Local Memory

Merge Conflict Handling

Selective Sync

📊 Benefits Summary

🚀 Next Steps

Ada Pertanyaan? Yuk Ngobrol!

🧠 Multi-Agent Shared Memory System

🎯 What You'll Build

🎯 Use Case Example

🏗️ Architecture Overview

📁 File Structure

🛠️ Step-by-Step Setup

Step 1: Create GitHub Repository

Step 2: Generate GitHub PAT (Personal Access Token)

Step 3: Create Core Memory Files

Step 4: Create Sync Script

Step 5: Push to GitHub

🤖 Agent Setup

Configure Each Agent

⏰ Automation with Cron

Auto-Sync Every 15 Minutes

Startup Sync

🔄 Sync Flow Diagram

✅ Verification

Test Sync

Check Git History

🎓 Advanced Features

Agent-Specific Local Memory

Merge Conflict Handling

Selective Sync

📊 Benefits Summary

🚀 Next Steps

📚 Related Tutorials

Ada Pertanyaan? Yuk Ngobrol!