Pricing & editions

Same brain.
Different body.

Three ways to deploy your AI office. Same 10+ agents, same self-improvement, same results. Pick the infrastructure that fits your business. Upgrade anytime without changing a single agent skill.

See editions Compare all 6 products
Three editions

Your AI office runs where you want it.

Enterprise sovereignty. Managed simplicity. Or minimum viable cost. The agents don't care which one you pick.

Self-hosted

On-Premise

Your hardware. Your data. Your rules.
$0 / inference
Hardware investment: DGX Spark (~$3-5K one-time)
  • NVIDIA DGX Spark with local Nemotron 120B
  • 80% of tasks at zero marginal cost
  • Data never leaves your building
  • Full kernel-level sandbox isolation
  • Hybrid option: add cloud AI for complex tasks
  • Scale: link multiple Spark units (400B, 700B)
IT.CAFE managed

Managed Office

We run it. You use it. Zero ops burden.
$299 / month
Includes infrastructure + local inference + support
  • Hosted on our infrastructure (EU data centers)
  • Local LLM via our DGX Spark fleet (no cloud API costs)
  • We handle updates, backups, monitoring
  • Your agents, your skills, your Treasure
  • Optional: cloud AI boost for peak demand
  • SLA with guaranteed uptime
Customer VPS

Cloud Start

Deploy in 5 minutes. Scale when ready.
$15 / month
VPS cost + LLM source of your choice
  • Any VPS provider (Hostinger, OVH, Hetzner, AWS)
  • 1-click deployment via Docker
  • Choose your LLM: our GPU, cloud API, or both
  • Same agents, same skills, same Treasure
  • 2 vCores, 4GB RAM minimum
  • Upgrade to Managed or Self-Hosted anytime
Mix and match

Pick your AI brain. Change it anytime.

Every edition works with every LLM source. Start with cloud API at $0 upfront. Switch to local inference when the math makes sense. Or run hybrid from day one.

Local GPU (yours)

Full Sovereignty

Nemotron 120B on your DGX Spark
Cost per task$0
Latency1-2s
Data privacyMaximum
Available onSelf-Hosted
Local GPU (ours)

Best Value

Our DGX Spark serves your agents
Cost per taskIncluded / low flat fee
Latency2-5s
Data privacyHigh (our servers, not cloud)
Available onManaged, VPS
Cloud API / Hybrid

Flexible

Claude, GPT, Gemini — or mix them
Cost per task$0.01-0.10
Latency3-8s
Data privacyProvider-dependent
Available onAll editions
The full picture

6 products. One upgrade path.

Every combination that makes sense. Start anywhere, upgrade without migration.

# Product Infrastructure AI Brain Monthly Cost Best For
01 SovereignFlagship Your DGX Spark Local Nemotron 120B ~$80-150* Enterprise, regulated industries, data sovereignty
02 Sovereign Hybrid Your DGX Spark 80% local + 20% cloud ~$130-350* Enterprise needing peak cloud power
03 Managed OfficeSweet spot Our infrastructure Our local GPU $299 SMB wanting AI office without ops
04 Managed Starter Our infrastructure Cloud API $149 + usage SMB starting fast, upgrade to GPU later
05 VPS + Our BrainBest value Your VPS Our local GPU $65-170 Tech-savvy SMB, cost-conscious
06 VPS Starter Your VPS Cloud API / Hybrid $15 + usage Solo business, startups, developers
* Hardware amortized over 36 months. Cloud API usage varies by volume.
The upgrade path

Start cheap. Upgrade without migration.

Your agents, skills, and learned improvements move with you. Only the infrastructure config changes.

Step 1

VPS Starter

$15/mo + cloud API. Prove the value. 5-minute deploy.

Step 2

VPS + Our Brain

Switch inference endpoint. Lower cost, better latency.

Step 3

Managed Office

We take over ops. You just use it.

Step 4

Sovereign

Buy DGX Spark. Full on-prem. Zero cloud dependency.

Each upgrade: change one config file. Agent skills don't change.
Treasure (6 months of learned improvements) migrates with a single export/import.
What's portable

The brain is the same everywhere.

Every edition runs identical agent skills. Your AI office doesn't care what infrastructure it sits on.

🏛

Agent Hierarchy

Patron, AZ, Barista + team. Governance, escalation, approval gates.

📈

Treasure Loop

Self-improvement engine. Gets better at your specific work every week.

🔧

Universal Bridge

Zoho, Odoo, Microsoft — agents call capabilities, not vendor APIs.

🛡

Compliance Built-in

ISO 27001 + 42001 + ITIL 4 practices from day one.

📋

P3.express PM

Real project management methodology. Not AI guessing.

🧠

10+ Agents

Barista, Router, Worker, Developer, Reporter, AZ, PMO, Compliance, Gardener, Agnes.

The team

10 agents. 7 departments. Always on.

☕
Barista
Front desk & coordinator
🎯
AZ
Quality gate & approvals
🔀
Router
Classifies & dispatches
💻
Worker
General task execution
⌨
Developer
Code, test & debug
📊
Reporter
Quality & daily summary
📋
PMO
Project management
🛡
Compliance
ISO & ITIL audits
🌱
Gardener
Self-improvement engine
📣
Agnes
Marketing & content
Ready?

Start with $15/month.
Scale to full sovereignty.

Same agents. Same skills. Same self-improvement. Just pick your starting point.

Book a demo Back to overview