We use only essential, cookie‑free logs by default. Turn on analytics to help us improve. Read our Privacy Policy.
Formerly Kenaz

AI systems that ship, stay secure, and keep your data yours.

Engineering, security, and sovereignty for production AI.

FIG. 01 — PERPETUAL DELIVERY LOOP

Quick Answers about 37lab

What does 37lab do?

We build, secure, and deploy production AI systems — custom agents, adversarial validation, sovereign infrastructure.

Who is 37lab for?

Engineering leaders at companies in regulated industries who need AI that works in production, not just in demos.

What makes 37lab different?

We're engineers, not resellers. We built Mantis (our AI security scanner) and deploy GDPR-native sovereign solutions.

What We Do

Three practices. One standard: production.

AI Engineering

Build me AI that works

Custom agents, RAG systems, MCP servers, multi-agent pipelines. We build production AI — not demos that die after the board meeting.

  • Custom AI agents with persistent memory
  • RAG & knowledge systems
  • MCP server integration
  • Multi-agent orchestration
AI Engineering →

AI Security

Break it before attackers do

Red teaming, adversarial validation, EU AI Act compliance. We test your AI the way real threats do — not the way benchmarks pretend to.

  • Red teaming & adversarial validation
  • EU AI Act compliance
  • AI safety & compliance audits
  • Powered by Mantis — our own scanner
AI Security →

Sovereign AI

Your data stays yours

On-premise deployment, air-gapped infrastructure, privacy architecture. For organizations where 'we take privacy seriously' needs to mean something.

  • On-premise & air-gapped deployment
  • Privacy architecture design
  • GDPR/HIPAA compliance engineering
  • Data sovereignty by design
Sovereign AI →

Products

Built for ourselves first. Running in production.

37lab Core

Sovereign AI Platform

The platform layer every 37lab system stands on: a semantic foundation, MCP bridges to your databases and APIs, working agents, human-gated workflows — on infrastructure you own. Watch it assemble, layer by layer.

  • Ontology first: a semantic map every agent shares
  • MCP bridges: databases, APIs, services and memory on one bus
  • Agents and workflows with human-in-the-loop gates
  • EU cloud, self-hosted or air-gapped — your keys, your jurisdiction
See the assembly →
APISSERVICESDATABASEMEMORYHITL GATEFIG. 03 — CORE · ASSEMBLY VIEWSOVEREIGN BY DESIGN

Mantis

AI Security Scanner

We built a tool that finds what others miss. Mantis runs autonomous security analysis on AI-powered codebases — prompt injection, MCP abuse, unsafe tool use, data exfiltration paths, and more.

  • AI-specific vulnerability detection: prompt injection, MCP abuse, unsafe tool use
  • Automated triage with LLM-powered false positive reduction
  • OWASP Top 10 for LLM Applications coverage
  • Actionable reports with severity scoring and remediation guidance
Learn more →
Ø 5601 STATIC ANALYSIS · 56 SCANNERS02 AI VALIDATION · −85% FALSE POS.03 RED TEAMING · 47 PLAYBOOKS04 AUDIT-READY REPORTFIG. 02 — MANTIS · FRONT ELEVATIONPREDATORY BY DESIGN

Why 37lab

Engineers, Not Resellers

We write the code. No white-labeled platforms, no outsourced teams. You work with the people who build your system.

Own Product, Own Standards

We built Mantis — an AI security scanner running in production. We don't just audit AI systems, we build them.

GDPR-Native, EU-Based

Not GDPR-compliant as a feature. GDPR-native by architecture. Your data stays in your infrastructure, under your jurisdiction.

Production, Not Demos

90 days from first conversation to operational AI. We ship. If we can't, we tell you before you spend money.

Ready to build?

Confidential consultation · NDA available

Start a project