StonewaveEN DECreate account

AI engineering

We buildyour AI.

We build our own apps with a team of people and agents: a voice assistant in Rust, an office suite, an AI gateway, test benches. We put the same skills to work for companies, from the first idea to a system in daily use.

01 · Services

Six things we do well.

Agents & automation

Agents that take routine work off people: sorting mail, preparing bookings, checking documents.

  • Approval steps before money or mail goes out
  • Names replaced before a model sees them
  • Always on, or only when you want

AI in your software

Language models built into your products and internal tools: search, summaries, assistants, extraction.

  • Model choice by measurement, not hype
  • Costs per task known up front
  • Works with your existing systems

In-house AI

AI on your own hardware when data must not leave the house. You buy the hardware, we set up and maintain the system.

  • Open models, run locally
  • Sized to your users and tasks
  • Setup, updates and maintenance

Apps & platforms

Web, desktop and mobile, from a clickable prototype to a product that can grow.

  • Desktop apps in Rust and Tauri
  • Web platforms with accounts and billing
  • Prototype first, then the real thing

Evaluation & test benches

How often is your AI right, and where does it gloss over? We build the bench that measures it, from our own research.

  • Test sets from your real cases
  • Repeatable runs, numbers you can check
  • Before and after every change

Workshops & advice

Where AI pays off in your company and where it does not. An honest assessment, even when the answer is “not yet”.

  • Half-day workshop with your team
  • Map of tasks and risks
  • Plan with first steps

02 · Proof

What we built for ourselves.

The best reference we can show today is our own work. Every app below was built by two founders and a team of agents.

Stonewave account & AI gateway

Passwordless sign-in, usage per plan, one gateway for all models.

Agent Authority Protocol

Open-source protocol for what an agent may do on whose behalf.

Research harness

Preregistered experiments on AI checking AI, public with data and code.

Technology we use daily

  • Rust
  • TypeScript
  • Tauri
  • React
  • Astro
  • Node.js
  • Python
  • Claude
  • OpenRouter
  • Open models
  • Speech in & out
  • CalDAV

03 · How a project runs

Four steps.No surprises.

  1. We talk.

    You tell us the problem. We tell you honestly whether we are the right team.

  2. We agree what “done” means.

    A short written scope, including how we measure success, and a price.

  3. We build in small steps.

    You see running software early and often, not a big reveal at the end.

  4. We show the result.

    With numbers and how they were measured. Agents draft and build; people review and approve.

Terms

Fixed price or time and material

Depending on how clear the scope is. You get the price in writing after the first conversation.

Your code, your data

What we build for you belongs to you. Data stays where you decide.

Remote, from Austria

We work remotely with companies in German and English.

Reference in preparation. We show client work here once the client has approved the text.

A project in mind?

Write us a few sentences. We answer personally, usually with one or two questions back.

Start a project