🏗️ System Architecture
┌─────────────────────────────────────────────────────────────────────────┐
│ 🌐 CLAUDE AGENT SWARM │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ 💻 WINDOWS CLAUDE 🖥️ VM1 LEADER ☁️ GOGA CLAUDE │
│ Local PC Oracle Cloud Google Cloud │
│ Claude Flow V3 tmux session tmux session │
│ 15 sub-agents 5 sub-agents 3 sub-agents │
│ │ │ │ │
│ └─────────────────────────┼───────────────────────┘ │
│ │ │
│ ▼ │
│ SHARED BRAIN │
│ mcp-hub-old-frost-2327.fly.dev/brain │
│ │
│ ┌────────────────────┐ │
│ │ Message Exchange │ │
│ │ Task Coordination │ │
│ │ Status Tracking │ │
│ └────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────┘
📅 Experiment Timeline
18:00 UTC
Claude Leader Setup
Created persistent Claude agent on VM1 with autonomy settings
18:15 UTC
Cloud Backups Created
VM1, VM2 (Oracle 47GB each), Goga (Google snapshot)
18:20 UTC
Windows MAX Upgrade
Enabled Claude Flow V3, cowork mode, extended thinking
18:25 UTC
VM1 Cowork Enabled
Updated settings with cowork: 5 sub-agents, extended thinking
18:30 UTC
Goga Experiment Started
SSH connection established, npm installed
18:35 UTC
Claude Code on Goga
Successfully installed Claude Code on Google Cloud VM
18:40 UTC
OAuth Completed
Goga Claude authorized with Max subscription
18:42 UTC
🎉 Swarm Online!
All 3 agents connected to Shared Brain, experiment successful