SetechStudio

Independent studio · games and software

Games that ship.
Systems you keep.

Setech Studio is a three-person team of senior engineers who direct AI coding agents. We make our own games and build apps and business systems to order. You get a working product, the code, the tests and the whole delivery pipeline, so your team can keep shipping without us.

Play in your browser on Setech Arcade
  • 4games you can play now or soon
  • 5 wksfor Neon City, one person
  • 3open-source tools
  • 3people on the team

Projects

Don’t take our word for it. Play it.

Every one of these games was built with the same process we offer our clients. Code, metrics and the making-of are public.

Playable demoBrowser · 3D · co-op

Neon City

A co-op 3D shooter set on the real squares of Polish cities.

  • 5 weeksof one developer’s time
  • 869commits in 15 iterations
  • 2 500+tests, all green
Demo coming soonBrowser · Windows · macOS · Linux

HEMS

A rescue helicopter over the real Poland.

  • 3 weeksof one developer’s time
  • 1:1terrain of the Łódź region
  • 2flight models, with autorotation
Browser demo coming soonDetails

Open-source tools

Our pipeline is not a black box. Its key parts are public.

v0.23 · MIT

agentctl

A task dispatcher for AI agents: spec, its own worktree, an agent, a report and a cross-model review. This is the “factory” a client can take over.

v0.3 alpha · MIT

archi-agent

Verifies software flows against your architecture model. Enterprise catalogues, C4 and PlantUML views, VS Code, CLI and MCP.

Live

Setech Arcade

Our Atari games in the browser: the atari800 emulator compiled to WebAssembly, disk images pinned by SHA-256, tests in Chromium, Firefox and WebKit.

How we work · Setech Agentic SDLC

Agents write the code. People decide.

AI agents write and test the code. People decide on architecture, scope and what goes to production. You see a working build every week. We also show you what didn’t work.

  1. 1Spec

    A person defines the task, acceptance criteria and the limits of the change.

  2. 2Agent

    The agent works on its own branch, with no access to production.

  3. 3Tests and scans

    Tests, visual regression, performance budgets, dependency and secret scans.

  4. 4Cross-model review

    A different model reviews the change than the one that wrote it.

  5. 5Human decision

    A person approves every merge.

Services

We build your system, then hand you the machine that built it.

Games, internal systems, apps, integrations and tools. We price outcomes, not hours. Prices exclude VAT; the final quote follows the Discovery Sprint.

01

Discovery Sprint

from €1,5003–5 days

Process map, architecture, a prototype of the riskiest part and a fixed quote. The fee is credited to the build.

02

Prototype or MVP

from €4,5002–8 weeks

A working build every week, code and CI in your repository, 30 days of fixes after launch.

03

Pipeline & handover

from €6,0002–4 weeks

We set up agents, CI/CD, tests and review gates in your repository, train your team and leave runbooks behind.

04

Ongoing development

from €3,500 / month3 months minimum

Reserved team capacity plus pipeline care.

You own everything

Code, cloud account, data, agent configuration and docs are yours from day one. No proprietary platform, no exit fee.

  • Code and full history in your repository
  • Agent instructions, skills, hooks and spec templates
  • CI/CD with test, review and performance gates
  • Runbooks and architecture decision records
  • Pairing days with your team
  • 30 days of support after handover

Team

Talk to the people who build it.

Three people, no sales layer, no account managers. We reply within one business day.

Marcin KrzetowskiFounder · architecture, games, agent pipeline
Coming soonSecond team member
Coming soonThird team member

FAQ

Common questions

Is AI-written code safe and good?

Every change goes through tests, dependency and secret scans and a review by a different model, and a person approves every merge. Agents have no access to production.

If it’s AI, why isn’t it cheaper?

We price outcomes, not hours. For the same budget you get more features, sooner. With a smaller budget we cut scope, never tests or review.

Who owns the code?

You do. The contract transfers copyright and describes the human contribution.

What if we stop working together?

Your team has the code, the pipeline, the agent configuration and the runbooks. Nothing runs only on our side.

What about data and GDPR?

We sign a data processing agreement and tell you which models we use. Client data is not used to train models.

Contact

Book a 30-minute call.

We’ll tell you honestly whether and how we can help. If the project isn’t a fit, you’ll hear that too.