Ciprian Comsa

Hi dev.to, I've been building Nodlyn, a visual workflow builder with one specific obsession: it auto-discovers devices — industrial PLCs, smart home gadgets, whatever's on the network — and lets you wire them into automations without knowing the protocol underneath.

The gap I kept hitting

Most no-code automation tools (Zapier, Make, n8n) are built for connecting SaaS apps to each other. That's a solved problem. What isn't solved: connecting a Siemens S7 PLC, a Modbus temperature sensor, or a Zigbee bulb to a workflow, without already knowing what a register address or a node ID is. The industrial tools that do speak those protocols (Ignition, and friends) assume you're a controls engineer. The IoT tools that are easy to use (Home Assistant) don't touch industrial protocols at all.

Nodlyn's core feature is a discovery panel: hit scan, and it finds what's on your network — Modbus device, Zigbee bulb, OPC UA server, whatever — and suggests a starting workflow for it. No manual protocol config, no address mapping by hand.

What's actually in it

  • Device discovery across 30+ protocols — MQTT, Modbus TCP, OPC UA, BACnet, Siemens S7, Matter, CAN Bus, ZigBee, Z-Wave, LoRaWAN, BLE, and more, all from the same discovery panel
  • 350+ connectors/blocks and 230+ templates, business-side stuff included: SFTP, webhooks, Stripe, Salesforce, email/IMAP triggers
  • Local LLM support (Ollama, LM Studio) alongside OpenAI-compatible cloud APIs — for teams that can't send documents to an external model
  • A ladder-diagram view, toggled from the same toolbar as the drag-and-drop canvas, for anyone with a PLC background who thinks in contacts and coils, not nodes and edges
  • Custom components in C# — drop a DLL in the plugins folder, it hot-reloads into the node library, no restart
  • Runs on-prem by default. The desktop app (Studio, Windows for now) builds workflows locally. You can export any workflow as a standalone runtime agent — headless, deployable to Windows, Linux, or macOS — that executes independently on a server, a factory PC, or a Raspberry Pi. Cloud monitoring (heartbeats, remote pause/resume, logs) is opt-in, not required.

What's rough right now

It's beta. The Studio desktop app is Windows-only so far (Linux/macOS builds for Studio itself are on the roadmap — runtime agents already support all three). The installer isn't code-signed yet, so Windows SmartScreen will warn you on first install ("More info" → "Run anyway" gets you through it) — a Microsoft Store submission is in review right now specifically to fix that.

What I'd love feedback on

Feedback is welcome, particularly from anyone working on edge/industrial automation or already running self-hosted workflow tools. Two specific questions: which protocol, if missing, would be a dealbreaker for your use case? And separately — what would you need to see demonstrated (not just described) to trust that device discovery works without manual protocol configuration?

Link: https://nodlyn.com — free beta, full feature set unlocked.

Happy to answer anything about the architecture (C# plugin system, how discovery works under the hood, the runtime agent design) in the comments.