AI · 2026

We put AI to work.

We do not manufacture pilots. We integrate and run. Claude and GPT in your product, or Llama 4 / Qwen3 on our own hardware. Agentic workflows with MCP, OpenClaw-based agents, AI security. In live systems, not demos.

Ask about an AI project AI security

What we do with AI

Three directions, one team.

LLM into product

RAG, copilots, document extraction, semantic search. Claude, GPT, or open-source — depending on what fits technically and commercially.

RAG over your own knowledge base
Copilots in admin panels, customer portals
Document reading, structuring, extraction
Semantic search where an old search engine sat

Sovereign / on-prem AI

Open-source models running on our GPU — or yours. Data does not leave the EU, and in most cases not even Hungary.

Llama 4 Scout/Maverick, Qwen3, DeepSeek deployment
Embedding models with multi-lingual (Hungarian!) support
vLLM / TGI clusters, edge proxy, rate limiting
GDPR-compliant data-flow documentation

Agentic workflows

Multi-step AI agents for internal processes. Claude/GPT or self-hosted, with MCP integrations and human checkpoints.

Multi-step agents (tool use, tool calling)
MCP servers for your systems
OpenClaw-based local-first agents
Internal ops automation: content, support, data
Human checkpoints at every critical step

What we use daily

Today's AI stack — what we actually ship.

We don't list every trend. Only what we've shipped in production or can deliver reliably next quarter. If we can't stand behind it in an interview, it isn't here.

Agentic systems: LangGraph, CrewAI, OpenClaw, custom orchestration. Only what we have actually shipped in production.
Local-first agents: OpenClaw deployment and customisation — personal and internal agents that run on your own machine or server, not in the cloud. Markdown-based memory, messenger UI, any LLM as the backend.
Coding agents: Claude Code, Cursor, Aider in our own flow. We know where they break — which is why we can bring them to client work responsibly.
MCP integrations: MCP servers for client systems so Claude/GPT/local models see them as native tools.
Open-source models: Llama 4 Scout/Maverick, Qwen3, DeepSeek deployment for clients with data-sovereignty constraints.

Rule: this list only contains things we could present with a reference in a technical interview. If a trendy technology is missing, it means we're not yet at a place where we'd credibly commit to a project built on it.

AI security

We don't just build — we defend.

Prompt injection testing — we red-team LLM apps before they see real traffic.
RAG governance — access control, PII redaction, audit trail across retrieval pipelines.
Model access control — who can call which model with which data; rate limits, budget killswitch.

Full security page

How it works

Problem in, solution out.

Got an AI idea? Start here.

A few questions become an AI draft our team reviews. Specific approach, stack picks, estimate band, risks.

Let's talk Services