I built a factory simulation with 30 machines (CNC, injection molding, assembly lines) that runs 24/7 with realistic production schedules, breakdowns, and quality issues.
On top of it sits an AI layer with 111 MCP tools.
So you can connect any LLM and query ERP, OEE, WMS, QMS, and scheduling data in real time.
The idea: Manufacturing AI is hard to learn because you never get access to a real factory. This gives you one — for free.
What you can do:
- Chat with a factory AI agent that has access to live production data
- Build visual AI workflows with a Node-RED-based flow editor
- Take challenges ("Get OEE above 85%", "Eliminate late deliveries")
- Bring your own LLM key (OpenAI, Anthropic, or local models)
Stack: Next.js, Express, Node-RED per-user pods on K8s, PostgreSQL, factory simulator in TypeScript. The MCP tools follow the Model Context Protocol spec.
tobiaslante•2h ago
On top of it sits an AI layer with 111 MCP tools.
So you can connect any LLM and query ERP, OEE, WMS, QMS, and scheduling data in real time.
The idea: Manufacturing AI is hard to learn because you never get access to a real factory. This gives you one — for free.
What you can do: - Chat with a factory AI agent that has access to live production data - Build visual AI workflows with a Node-RED-based flow editor - Take challenges ("Get OEE above 85%", "Eliminate late deliveries") - Bring your own LLM key (OpenAI, Anthropic, or local models)
Stack: Next.js, Express, Node-RED per-user pods on K8s, PostgreSQL, factory simulator in TypeScript. The MCP tools follow the Model Context Protocol spec.
Source: https://github.com/TobiasLante/openshopfloor
Would love feedback — especially from anyone working in manufacturing or building with MCP tools.