Research

FounderStack

A nine-layer reference architecture for contract-driven, swappable AI agents.

01 / Problem

The problem.

Solo founders adopt AI agents fast, but without an architecture they end up locked in to one vendor and hard to trust.

02 / Solution

The solution.

A nine-layer, contract-driven reference architecture where agents, models and skills can be swapped by manifest. Evaluated live on HempFinder.

03 / Principles

The Neotron way.

Sovereign(applied)
Local AI(applied)
EU / GDPR-first(not yet)Not yet
Open source(applied)
Circular(not yet)Not yet

Agents run on local models by default, every layer is swappable, and humans keep the approval gates.

04 / Built with

Built with.

Backend

AI

gpt-oss-120b (local)
MCP

Runs on Neotron building blocks

Local AI server
Agent stack

05 / Impact

Impact.

AIS26 paper
A Swappable, Local-First Agent Stack, AI Summit Styria 2026

8
Decent work
Enabling
Lets solo founders do more without losing control of their tools.
9
Industry & innovation
Direct
Open reference architecture for trustworthy AI in small companies.

Portfolio impact map →

06 / Trend & innovation

Trend & innovation.

Technology readiness

TRL 4 of 9

Horizon

H2 — Emerging · 1–3 years

Megatrends

Konnektivität
Future of Work

Why now: Solo founders are adopting AI agents without an architecture that keeps them swappable and accountable.

Megatrends after the Zukunftsinstitut model.

07 / Get involved

Get involved.

Support

Back the work as a partner, pilot or early user.

Know someone who should see this?