MergenHub An in-app assistant that knows your plant
Talk to ChatGPT, Claude, OpenRouter or a fully local LLM from inside MergenHub. The assistant acts with your permissions: it diagnoses devices, sets up alarms and log groups, designs SCADA screens, configures HTTPS and the firewall — showing every step and asking for consent.
What it does
The assistant uses MergenHub's 100+ tools (channels/devices/tags, live values, alarms, historian, SCADA, settings, users, network tests) with the signed-in user's permissions. "Find devices that are not communicating", "create a level alarm for this tank", "draw a pump line screen and preview it", "publish HTTPS on 443 for my domain" — it works step by step like an automation engineer and verifies the result.
Key features
25 provider presets (cloud + local), searchable model picker, local server scan, reasoning/thinking mode, image and file attachments (vision), clickable channel/device/tag references, context meter with automatic summarisation, conversation folders and archive, site-specific skill packs, external MCP servers. There is never a tool that writes live values to a device.
Core capabilities
Cloud + local
OpenAI, Anthropic, OpenRouter, Gemini, Mistral, DeepSeek, Groq, xAI, Azure… and Ollama, LM Studio, vLLM, llama.cpp — keys stay on the server.
The user's permissions
No token: the assistant is the signed-in user. Write tools depend on the admin gate and the user's own permissions; every write is audited.
Answers by measuring
Device status, ping, TCP port checks, DNS, local ports; alarms, historian and event log — measurement, not guesswork.
Expert design
Real symbol library, pipes/flow animation, tank-pump scenes, ISA-101 colour language, validation by preview.
Settings & HTTPS
Partial settings updates, self-signed / PFX certificates, firewall ports, service restart — step by step, with consent.
Claude Code style
Status line ("Creating screen…"), tool cards, clickable references, attachments, thinking block, context meter, skills.