What I Actually Run
This isn't theoretical. Here's the infrastructure behind my personal AI team.
Agents
- →Gina — Claude Opus, WhatsApp + Telegram
- →Luna — Sonnet + Opus escalation, Telegram + Discord
- →6 specialized operators for dev, ops, research, comms, calendar, and orchestration
Infrastructure
- →OpenClaw on WSL2 (primary workstation)
- →DigitalOcean VPS for remote services
- →WireGuard VPN for secure access
- →Docker containers — Qdrant (vector DB), Firecrawl (web scraping)
Security
- →Daily automated security audits at 9:30 AM
- →Docker sandboxing for sub-agents
- →Allowlisted messaging — no open DMs
- →Credential isolation — secrets in .env only
Integrations
- →Google Workspace (Gmail, Calendar, Drive)
- →Asana, GitHub, GitLab, Cloudflare
- →ElevenLabs (TTS), Canva (design)
- →Custom CLI tools for places, weather, PDF editing
Cost
- →~$15/day API costs
- →$6/mo VPS (DigitalOcean)
- →Model routing: Haiku (volume), Sonnet (skilled work), Opus (strategy)
- →Total: under $500/mo for a full AI team