← back
live on App Store · Dec 2025 → Co-founder TEF Ignition Swift · C++ · Python

Plynx.

An AI workbench for hardware makers. One project unifies dashboard, firmware, wiring, PCB, 3D case and phone control — so an AI agent can synthesise an entire connected device end-to-end from a single prompt. We call it vibe making.

Daniele Lin and Niccolò Pagano on stage at Tech Europe Foundation Ignition pitching Plynx
Niccolò Pagano presenting the Plynx technical pitch
Tech Europe Foundation audience
80+
weekly active
6 vers
shipped in 4 mo
8
board families
32 B
params · fine-tuned

The technical bet.

Fine-tuning Qwen2.5-Coder-32B on agentic hardware-design tasks.

A 32-billion-parameter open-weights LLM fine-tuned for agentic project generation across the full Plynx workbench (dashboard, firmware, wiring, PCB, 3D). We built a derivative-data pipeline over open-source hardware projects to generate training signal. This is the core technical direction of the company — well outside undergraduate scope.

What's been built.

Canvas Studio — drag-and-drop dashboard editor (buttons, toggles, sliders, gauges, charts, RGB, terminals, tables). Code Studio — Arduino editor + compiler with in-editor AI completion. Wiring editor ↔ PCB editor — interconnected, sharing one project context (the core unlock for end-to-end AI generation). 3D editor — generates project enclosures. Marketplace with reviews, payments, customizable rules. Android app on the current protocol. Eventoring rule engine — conditions, actions, notifications, schedules, scenes. MQTT broker + Home Assistant integration for interop. Web platform — forum, documentation, public dashboard, "Scratch-for-Arduino" learning tool.

Custom binary wire protocol.

Designed from scratch during Ignition — binary on the wire, TCP/SSL transport, OTA firmware updates — and rolled out across server, iOS, Android, and a C++ firmware SDK. Replaces an earlier protocol; supports eight board families simultaneously with mixed-board projects as first-class citizens.

Arduino Uno/MegaESP8266ESP32 Raspberry PiTeensySTM32Particle

Model Context Protocol (MCP) for IoT.

Plynx exposes its devices over Anthropic's MCP, so any MCP-compatible AI agent can speak to real hardware through Plynx. Early adopter of the MCP standard outside its native use case (developer tooling).

Backing.

Tech Europe Foundation Ignition cohort (Bocconi, Politecnico di Milano, ION). €2 000 grant secured.


Co-founded with Niccolò Pagano. My focus inside the company: server & protocol design, firmware SDK, dashboard / editor implementation, MQTT / Home Assistant integration, MCP layer.