Sovereign AI Stack

Local-first multi-agent system. Your data, your models, your rules.

No cloud API dependencies. No subscription. No one else’s uptime requirement.

---

What this is

Sovereign AI Stack is a production-grade agent runtime you own end to end.

It runs on your hardware, talks to local LLMs through Ollama, discovers tools dynamically via MCP, coordinates multiple agents through an auction-based swarm, and even accepts Bitcoin micropayments for API access via L402.

You ship it behind your firewall, your VPN, or your home router. Your prompts, your memory, your data never leave your box.

---

Stack

• AI runtime: MCP tool discovery + Agno-style coordinator/sub-agent swarm + cascade LLM routing
• Interface: FastAPI backend + HTMX dashboard + WebSocket live feed
• Storage: SQLite with backup/restore runbooks
• Auth / revenue: L402 Lightning payment proxy (optional)
• Infra: Docker Compose deploy, GitHub Actions CI, 228+ tests, 60% coverage floor, watchdog monitoring
• Voice: Optional persona runtime with TTS/STT and RTMP ingest

---

Who this is for

• Privacy-first developers who want AI without sending prompts to OpenAI/Anthropic
• Bitcoin/sovereign operators running home labs and full nodes
• Engineering leads who need an internal agent runtime they can audit and extend
• Hackers/indie devs who want a local agent stack they can modify, not rent

---

What you get

Self-hosted

• Full Docker Compose deploy
• README + architecture diagram + threat model
• 30-minute setup on x86_64 or Apple Silicon
• Includes: MCP auto-discovery, coordinator swarm, dashboard, WebSocket UI, L402 payment gateway

Supported

• Everything in self-hosted
• 1-hour onboarding call
• 30-day support window
• Security patch notifications
• $199/month or $1,999/year

Enterprise

• Multi-node orchestration
• SSO + team auth
• Custom MCP tool integration
• On-prem install or air-gapped bundle
• $5,000 flat + support retainer

---

Architecture

┌─────────────────────────────────────────┐
│  Dashboard + WebSocket UI (HTMX)        │
├─────────────────────────────────────────┤
│  Coordinator / Swarm Router             │
│  Auction-based task assignment           │
├─────────────────────────────────────────┤
│  MCP Auto-Discovery Layer               │
│  Dynamic tool registration + execution  │
├─────────────────────────────────────────┤
│  Cascade LLM Router                     │
│  Ollama primary → provider fallback      │
├─────────────────────────────────────────┤
│  SQLite Persistence + Backup/Restore    │
├─────────────────────────────────────────┤
│  L402 Lightning Payment Proxy           │
│  Macaroon auth + Bitcoin micropayments   │
└─────────────────────────────────────────┘
         ↓ optional persona runtime
┌─────────────────────────────────────────┐
│  TTS / STT / RTMP / Avatar              │
└─────────────────────────────────────────┘

---

Security model

• All inference happens locally. No prompt exfiltration path to third-party APIs in default config.
• L402 uses HMAC macaroons, not credential databases.
• All state is SQLite-backed; backup/restore is tested.
• CI runs 228+ tests with coverage floor enforced by watchdog.
• No telemetry, no phone-home, no usage tracking by default.

---

Demo

Coming soon — 3-minute walkthrough of local-first agent completing a real task without cloud APIs.

---

Pricing

• Self-hosted: $2,500 one-time setup fee + optional support
• Supported: $199/month
• Enterprise: $5,000 flat + retainer

Payment accepted in USD or Lightning.

---

Contact

alex@alexanderwhitestone.com github.com/Alexspayne