Fleet Setup Guide
Benchmarked against: Anthropic — (No direct equivalent — SuperPortia SOP) Scope: Setting up a new ship from scratch Naming: SS{N} = SuperPortia Ship {N}, codename Beagle {N}
This guide covers how to set up a new ship in the SuperPortia fleet — from hardware to fully operational agent with UB access.
Fleet overview
| Ship | Platform | Location | Status |
|---|---|---|---|
| SS1 | Mac Mini M4 Pro | Office | Running |
| SS2 | Windows PC | Home | Building |
| SS3 | Cloudflare Workers | Cloud | Planning |
Setup sequence
Building a ship follows this order (AI Agent architect thinking, not traditional engineer thinking):
Key principle: Brain first, not repo first. UB is the agent's memory — without it, the agent is amnesiac.
Step 1: Device setup
Hardware requirements
| Component | Minimum | Recommended |
|---|---|---|
| CPU | 4 cores | 8+ cores |
| RAM | 8 GB | 16+ GB |
| Storage | 256 GB SSD | 512+ GB SSD |
| Network | Stable internet | Low latency |
Software prerequisites
| Software | Purpose | Required |
|---|---|---|
| Python 3.11+ | LangGraph line, UBI | Yes |
| Node.js 18+ | Docusaurus, MCP servers | Yes |
| Git | Version control | Yes |
| Claude Desktop App or CLI | Agent runtime | Yes |
Step 2: Brain setup (Cloud UB)
All ships connect to the same Cloud UB — no local UB setup needed.
Configure environment variables
# Ship identity
export SP_SHIP_ID="SS2" # Ship identifier
export SP_AGENT_ID="win-cli" # Agent identifier
# Cloud UB connection
export SP_CLOUD_UB_URL="https://ub.superportia.dev"
export SP_API_KEY="your-api-key"
Verify connection
# Test Cloud UB access
cloud_ub_health()
search_brain(query="test connection")
Step 3: Agent setup
Install Claude Code
Follow the official Claude Code installation for your platform.
Configure MCP servers
Create the MCP configuration with servers appropriate for the ship:
| Server | Required | Notes |
|---|---|---|
| Cloud UB | Yes | All ships need UB access |
| Local UBI | Optional | For local file operations |
| Cloudflare | Optional | For infrastructure management |
| Supabase | Optional | For database operations |
| Chrome | Optional | For browser automation |
| Memory | Recommended | For personal agent memory |
| Filesystem | Recommended | For file operations |
Configure hooks
Set up the SessionStart hook for boot sequence:
{
"hooks": {
"SessionStart": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "bash .claude/hooks/session-start.sh"
}
]
}
]
}
}
Configure rules
Copy the standard rule files to .claude/rules/:
| Rule file | Purpose |
|---|---|
company-constitution.md | Behavioral standards |
cost-awareness.md | Token discipline |
agent-intelligence-protocol.md | Mandatory behaviors |
tech-freshness.md | PK defense |
ub-governance.md | UB standards |
gsta-alignment.md | Strategic alignment |
compaction-recovery-protocol.md | Memory cliff defense |
glossary.md | Term navigation |
Step 4: Verification checklist
After setup, verify everything works:
| Check | Command | Expected |
|---|---|---|
| Agent online | agent_heartbeat() | Success response |
| Mailbox works | check_agent_mailbox() | Message list (may be empty) |
| UB search | search_brain("test") | Results from Cloud UB |
| UB ingest | ingest_fragment(path="Test entry", input_type="text") | Entry ID returned |
| WO access | list_work_orders() | WO list |
| Factory floor | factory_floor_status() | Shows all agents |
Ship-specific considerations
SS1 (Mac)
| Feature | Details |
|---|---|
| Shell | zsh (default on macOS) |
| Python | .venv for LangGraph ecosystem |
| Local UBI | Available (SQLite + Qdrant) |
| Photos Library | Accessible via SQLite |
| Chrome MCP | Available |
SS2 (Windows)
| Feature | Details |
|---|---|
| Shell | cmd / powershell |
| Python | .venv (different path format) |
| Local UBI | Needs setup |
| File paths | C:\Users\... (backslashes) |
| Chrome MCP | Available |
SS3 (Cloudflare)
| Feature | Details |
|---|---|
| Runtime | Workers (serverless) |
| No local filesystem | All storage via D1/R2/KV |
| No shell access | Operations via API only |
| Cloud UB | Native (runs on same infrastructure) |
Dual-machine monitoring
Ships should monitor each other for resilience:
Principle: Dual-machine mutual monitoring, not single-direction watchdog. Each ship checks the other.
Onboarding a new agent
When a new agent joins the fleet, it must know:
- Company Constitution — behavioral standards
- WO System — how to receive and complete tasks
- Ingestion Rules — how to add knowledge to UB
- Cost Awareness — engine selection discipline
- UB Access — how to search and retrieve knowledge
- Role-appropriate tools — which MCP servers to connect
All onboarding information comes from Cloud UB, not local files (Captain directive: "don't check local").
Related pages
| Page | Relationship |
|---|---|
| Setup MCP | MCP configuration details |
| First Heartbeat | Agent registration |
| Session Start | Boot sequence |
| Fleet Deployment | Fleet operations |