/ai

Private AI,
on your own server.

Self-hosted LLMs with RAG over your documents. No OpenAI, no data leaving the EU. We set it up, host it, and keep it running.

Start a briefAll services
eu-only · gdpr-native · source on handover

pricing

Three tiers. Pick the closest.

Every quote is written, not auto-generated. The closer your idea is to a tier, the faster we ship.

Starter

from 800€+ 199€/mo

Setup in 3-5 days

  • Self-hosted LLM (Llama 3.1 8B or equivalent)
  • Single Hetzner server, EU-based
  • Branded chat UI on your subdomain
  • GDPR-compliant by default
  • Email support, business hours
Start a brief
Most picked

Standard

from 1500€+ 499€/mo

Setup in 5-7 days

  • Mid-size LLM (Qwen 3 30B or Mixtral)
  • Custom RAG over your documents
  • Branded chat + admin dashboard
  • System-prompt tuned to your domain
  • Priority support + monthly review
Start a brief

Custom

from 3000€+ 800€/mo

2-4 weeks

  • Multi-tenant or fine-tuned model
  • Multi-domain RAG with citations
  • SSO / SAML integration
  • SLA, on-call, audit logs
  • Compliance docs (AVV, ToM, DPA)
Start a brief

Or describe what you need — custom quote in 24h.

what you get

No surveillance. Just useful AI.

Self-hosted, on your perimeter

Your model runs on your hardware or our EU servers. Nothing leaves the boundary you set. No OpenAI, no Anthropic, no US cloud.

GDPR-native, not an afterthought

Built for Art. 32 TOMs from day one. AVV, DPA, retention rules, audit logs — included, signed, ready for your DPO.

Custom RAG over your docs

We ingest your knowledge base, build the vector store, tune the retrieval pipeline. Answers come with citations to the source paragraph.

Branded chat UI

White-label chat matched to your brand. Embed via iframe, subdomain, or full app. Your logo, your colors, your tone.

Monitored & observable

Live status page, alerting on latency and errors, monthly health reports. You see what the model sees.

Open weights, no lock-in

Llama 3.1, Qwen, Mistral, Mixtral — we pick what fits your workload. Swap models any time. No vendor pricing games.

how it works

Brief, quote, pay, ship.

  1. 01

    Brief

    Drop the idea in Discord or by mail. Rough scope is enough, refs welcome.

  2. 02

    Quote

    Fixed price + delivery window back within 24 hours. No hourly surprises.

  3. 03

    Pay

    50% upfront. Stripe, SEPA, Klarna, PayPal — or crypto (ETH, SOL, BTC). The rest on delivery.

  4. 04

    Ship

    48h for small jobs, up to a week for full stacks. Source handed over with invoice.

faq

Private AI, answered.

Which models do you run?

Open-weight models only — Llama 3.1 (8B / 70B), Qwen 3 (8B / 30B), Mistral, Mixtral. We pick based on your workload, latency target, and hardware budget. Swap anytime.

Where does the server live?

Hetzner in Falkenstein or Nürnberg (Germany / EU) by default. On-prem deployment to your own datacenter is available on the Custom tier.

How is this different from ChatGPT or Claude?

Both ChatGPT and Claude send your prompts to US providers. We don't. Your data, your model, your server — fully under your control. You also pay a fixed monthly fee instead of per-token pricing that scales with usage.

How long from brief to live?

Starter goes live in 3-5 days. Standard with custom RAG ships in 5-7 days. Custom (multi-tenant, fine-tuned, SSO) takes 2-4 weeks depending on scope.