Xreative Labs

The AI arm of Xreative · The Netherlands

Systems that produce by themselves.

Xreative Labs builds AI systems that don't stop at the demo: editorial sites that publish every day, a search engine across 3.6 million court rulings, and the infrastructure that keeps it affordable. On our own hardware, behind our own model router.

The register

Not a portfolio carousel. This is what runs, and what state it is in.

Neuralex
Legal search platform across 3.6 million court rulings — semantically searchable, built around the question a citizen asks rather than the case number a lawyer knows.
Live · platform
Ultranova
Astronomy desk publishing a new object page every day, including the daily NASA image. Edited entirely by agents, with mandatory AI disclosure.
Live · editorial
Neuristics
Opinion and tech site with nine editorial pen names on an edge worker. Every piece carries an explicit AI statement, per the AI Act.
Live · editorial
llmnet
Content stack across ten subdomains with a daily cycle that throttles itself when the quota runs low. Publishing stops before it gets expensive.
Live · content
AgentLedger
Cost and governance layer showing consumption per agent, per model and per call. Built because “it's probably fine” is not a budget line.
Internal · infra
Sentinel
Router with automatic failover across five providers. If one drops out, neither the user nor the invoice notices.
Internal · infra

What people call us for

Editorial systems

Sites that publish daily without anyone pressing “post” — with sourcing, disclosure, and a brake that engages when quality drops.

Search across large corpora

Making millions of documents searchable in the user's own words. Retrieval that explains why something ranks first.

Agent infrastructure

Routers, cost accounting, failover and memory. The invisible layer that decides whether an AI product is still viable in six months.

“Anyone can build a demo. The question is what happens on day ninety, when nobody is watching.”
  1. Measure before we buildEvery proposal starts with a number: what does this cost today, in hours or in tokens.
  2. Own hardware firstCritical services run natively on hardware we control, not only in a container that can fall over.
  3. Backup before every changeNothing is touched without a timestamped copy. Data is sacred.
  4. Failover by defaultOne provider is a single point of failure. Every route has a second and a third.
  5. Ship it transferableDocumentation, register and log book are part of delivery — not an attachment afterwards.

Contact

A system in mind, or an invoice climbing too fast?

Send two paragraphs on what needs to happen. You'll get an honest answer on whether it can be built — and whether it's worth it.