The Industrial AI that does the work.

AI for industrial applications means using AI to work with industrial data, equipment and processes. Common uses include helping engineers build automation applications, analysing machine data, inspecting products with computer vision, and allowing people or AI agents to interact with production systems. FlowFuse provides a platform for connecting these AI capabilities to industrial data and applications.

AI can support different parts of an industrial operation, from engineering and quality inspection to production monitoring and automation. For example, it can help generate application logic, analyse live machine data, identify defects from camera feeds, or let an AI agent interact with industrial systems. FlowFuse brings these capabilities into the same environment as the industrial applications and data they need to work with.

Examples include AI-assisted engineering, automated visual inspection, natural-language access to production data, AI agents that interact with industrial systems, and AI models running locally at the edge. With FlowFuse, teams can use FlowFuse Expert to build and explain applications, connect external AI agents over MCP, run ONNX vision models, and use cloud or local LLMs within industrial workflows.

Most industrial AI analyses a copy of your data and hands you a recommendation, and a person still has to go and make the change. FlowFuse AI works inside the Node-RED flows already connected to your machines, so what it produces is the change itself: built, reviewed by you, and shipped to every site that needs it. FlowFuse sells no hardware and works with any vendor's equipment and any AI model.

Yes. FlowFuse is built on Node-RED, which connects to OPC UA, Modbus, EtherNet/IP, Siemens S7, MQTT and more, plus over 5,000 community nodes. It is designed for brownfield sites, so it works alongside your existing PLCs, MES, ERP and SCADA rather than replacing them, and anything FlowFuse can reach, the AI can work with.

Yes. FlowFuse acts as an MCP server, so Microsoft Copilot, ChatGPT, Claude or a local model can connect to it and work your platform. You add one address, sign in, and choose which teams the agent may act on and whether it gets Read, Write or Destructive access in each. Because the agent is yours, so is the model behind it.

Yes. Command-line and editor agents such as Cursor, Visual Studio Code, Gemini CLI, Claude Code and Codex connect to FlowFuse over MCP, using the same URL as any other agent. On a self-hosted platform, use your own FlowFuse domain instead. Setup for each client is in the third-party agents documentation.

Every AI action runs through the permissions you grant, checked on every call: Read, Write or Destructive, set team by team, so a call that reaches past its grant is refused. By default FlowFuse Expert shows you each change and waits for your approval before it writes or deletes anything. Nothing goes live on its own unless a team owner turns on agent deploy, and every action is recorded in the audit log, marked as coming from Expert or from your own agent.

Some FlowFuse Expert features use the OpenAI API, so the context you choose to share, such as a flow, your palette or debug logs, is sent to OpenAI to answer the request. No FlowFuse data is used to train models, and OpenAI deletes API data after 30 days of abuse monitoring. See data privacy for details.

Yes, for the AI that runs in your flows. You can self-host FlowFuse on your own infrastructure, including air-gapped sites, run ONNX vision models at the edge, and use the Ollama node against a local model with no API key. You can also connect an agent that runs on your own model. FlowFuse Expert itself uses a hosted model, so it needs an outbound connection.